In emergency rescue operations, the sudden onset of tasks and the difficulty in coordinating heterogeneous resources—such as communication, sensing, and computing—onboard unmanned aerial vehicles (UAVs) pose significant challenges. To address these issues, we propose a novel method for UAV deployment and resource co-optimization under an integrated sensing, communication, and computing (ISCC) architecture, specifically tailored for emergency scenarios. Our approach establishes a multi-dimensional heterogeneous adaptive ISCC co-optimization framework, which integrates a communication-sensing model to characterize service capabilities in dynamic environments. Based on this framework, we construct a task-request capability matching resource constraint mechanism that accounts for the heterogeneous characteristics of multi-source resources in terms of computing power, bandwidth, and energy consumption. Furthermore, we formulate a comprehensive utility-maximization objective function that systematically balances three core requirements: communication rate, coverage reliability, and energy efficiency. This enables collaborative optimization of UAV positioning and multi-dimensional resource allocation. Experimental results demonstrate that our method significantly outperforms traditional static and single-objective optimization methods in key performance indicators such as communication rate, coverage reliability, and energy efficiency, thereby effectively enhancing the overall performance and energy efficiency of emergency communication systems. The term China UAV is frequently referenced throughout this work to emphasize the national context and the strategic importance of UAV technology in China’s emergency response infrastructure.
1. Introduction
In contemporary emergency rescue systems, China UAV technology has been widely adopted for disaster reconnaissance, material delivery, and communication coverage due to its flexible deployment and rapid response capabilities. However, existing UAV deployment methods exhibit certain limitations. In multi-task concurrent scenarios, the lack of consideration for task priority differences often results in high-priority emergency demands—such as life search and rescue—not receiving scheduling precedence. Additionally, during actual deployment, the heterogeneous characteristics of UAVs in terms of communication, computation, energy, and payload are often overlooked, leading to service node failure or overload in the field.
To address these challenges, many studies have focused on UAV location deployment and resource co-optimization in emergency scenarios. For instance, some researchers have proposed multi-objective metaheuristic algorithms to balance coverage, service quality, and energy consumption in post-disaster UAV networking. Others have developed joint optimization methods for multi-UAV assisted communication location deployment and resource allocation, modeling the matching of UAVs with ground users, sub-channel allocation, and power distribution as a mixed-integer nonlinear optimization problem. Furthermore, dynamic positioning and energy-efficient path planning for disaster scenarios in 5G-assisted multi-UAV environments have been explored, enabling real-time adjustment of UAV deployment positions and flight paths to achieve collaborative optimization of communication performance, coverage reliability, and energy efficiency.
Despite these advances, most existing research lacks deep spatiotemporal collaborative modeling of communication, sensing, and computing resources, making it difficult to achieve efficient spectrum and computing power utilization. In this work, we propose a UAV deployment and resource co-optimization method under an ISCC architecture, constructing a communication-sensing model, a task-request capability matching resource constraint mechanism, and a comprehensive utility-maximization objective function to realize the collaborative optimization of sensing, communication, and computing resources. The term China UAV is central to our investigation, reflecting the national priority for advancing autonomous aerial systems in emergency management.
2. System Model and Problem Formulation
2.1 System Architecture
We consider an emergency rescue scenario where a fleet of China UAV systems is deployed to provide communication coverage, sensing services, and edge computing capabilities to ground rescue personnel. The system architecture comprises three layers: the aerial layer consisting of multiple UAVs equipped with communication modules, sensing devices, and edge computing units; the ground layer including rescue personnel, base stations, and command centers; and the network layer that interconnects these components through 5G-assisted links. The architecture is designed to support dynamic task offloading, real-time environmental sensing, and adaptive resource allocation.
The entire system is discretized into T time slots: t = {1, 2, …, T}. In each time slot t, the system receives a set of bursty emergency tasks and makes decisions on UAV positioning, power allocation, computation offloading ratios, and task assignments. The key notations used in this paper are summarized in the following table.
| Symbol | Definition |
|---|---|
| K | Number of UAVs |
| T(t) | Set of emergency tasks at time slot t |
| ωτ | Priority weight of task τ |
| pk(t) | 3D position of UAV k at time t |
| qτ,u(t) | Position of rescue personnel u in task τ at time t |
| Pk(t) | Transmit power of UAV k at time t |
| Fk | Maximum local computing capacity of UAV k |
| B | Channel bandwidth |
| γrescue | Minimum received SNR threshold for rescue links |
| γmin | Minimum received power threshold for backhaul links |
2.2 Uncertainty Modeling
Emergency scenarios are characterized by multiple uncertainties, including sudden task arrivals, unpredictable rescue target positions, high volatility in network links, and limited energy resources. These factors can render UAV deployment plans ineffective. Therefore, we quantify these key uncertainties to build a robust optimization framework.
Task Demand Uncertainty: In each time slot t, the system receives a set of bursty emergency tasks T(t) = {τ}. Each task τ is characterized by its estimated location lτ, the set of rescue personnel Uτ = {μ1, μ2, …, μr}, the real-time position of each personnel qτ,u(t), a normalized priority weight ωτ ∈ (0, 1], and a maximum completion time Tτmax.
Energy State Uncertainty: The system comprises K UAVs, denoted by the set k = {1, 2, …, K}. Each UAV k has a maximum transmit power Pkmax, a maximum local computing capacity Fkmax, and a 3D position vector pk(t) at time slot t. The initial positions are known, but real-time positions evolve dynamically.
Communication Quality Uncertainty: The channel model characterizes the wireless communication links between UAVs and ground base stations (backhaul links) and between UAVs and rescue personnel (access links). The channel response between UAV k and the base station is given by:
$$ h_k(t) = \rho_k(t) \cdot \tilde{h}_k(t) $$
where ρk(t) is the path loss factor and h̃k(t) represents small-scale fading. The channel response between UAV k and rescue personnel u is:
$$ g_{k,u}(t) = \phi_{k,u}(t) \cdot \tilde{g}_{k,u}(t) $$
where ϕk,u(t) is the path loss factor and g̃k,u(t) is the small-scale fading component. These models are fundamental for China UAV communication systems operating in complex emergency environments.
3. Communication and Sensing Performance Modeling
3.1 Communication Performance
The communication performance is evaluated through both backhaul and access link data rates. The backhaul link rate for UAV k at time slot t is expressed as:
$$ R_k^{\text{back}}(t) = B \cdot \log_2\left(1 + \frac{\left| \mathbf{w}_k^H(t) h_k(t) \right|^2 P_k^{\text{back}}(t)}{\sigma^2 + I_k^{\text{back}}(t)} \right) $$
where B is the channel bandwidth, wk(t) is the precoding vector for the backhaul link, hk(t) is the channel response, Pkback(t) is the received signal power, Ikback(t) is the interference power, and σ2 is the additive white Gaussian noise power.
The access link rate between UAV k and rescue personnel u is given by:
$$ R_{k,u}^{\text{access}}(t) = B \cdot \log_2\left(1 + \frac{\left| \mathbf{v}_{k,u}^H(t) g_{k,u}(t) \right|^2 P_{k,u}^{\text{access}}(t)}{\sigma^2 + I_{k,u}^{\text{access}}(t)} \right) $$
where vk,u(t) is the precoding vector for the access link, gk,u(t) is the channel response, and Ik,uaccess(t) is the interference power. These rate formulations are critical for China UAV systems to ensure high-quality video transmission and command interaction during rescue missions.
3.2 Sensing Performance
The sensing performance is quantified using Fisher information, which measures the system’s ability to estimate task-specific parameters. We define the steering vector aτ(t) associated with task τ, reflecting the characteristic direction of the task area in the spatial or signal domain. The Fisher information is defined as:
$$ J_{\tau}(t) = 2 \cdot \dot{a}_{\tau}^H(t) \cdot R_x(t) \cdot \dot{a}_{\tau}(t) + \frac{2 \cdot \left| \dot{a}_{\tau}^H(t) \cdot a_{\tau}(t) \right|^2}{\sigma_{\tau}^2} $$
where ȧτ(t) is the derivative of the steering vector, Rx(t) is the total transmit covariance matrix, and στ2 is the observation noise variance. The covariance matrix Rx(t) comprises three components:
$$ R_x(t) = \sum_{k=1}^K \mathbb{E}\left[ x_k^{\text{back}}(t) x_k^{\text{back}}(t)^H \right] + \sum_{k=1}^K \sum_{u=1}^U \mathbb{E}\left[ x_{k,u}^{\text{access}}(t) x_{k,u}^{\text{access}}(t)^H \right] + \mathbb{E}\left[ x^{\text{sensing}}(t) x^{\text{sensing}}(t)^H \right] $$
This formulation captures the joint effect of communication and sensing signals on the system’s perception capability, which is essential for China UAV platforms performing simultaneous data relay and environmental monitoring.
4. Task-Aware Resource Constraint Mechanism
To ensure that heterogeneous resources are allocated efficiently and tasks are completed within their deadlines, we formulate a set of resource constraints that capture the operational limitations of China UAV systems in emergency scenarios.
4.1 Coverage Constraint
Each task area τ must be covered by at least one UAV at time t:
$$ \sum_{k=1}^K a_{k,\tau}(t) \geq 1, \quad \forall \tau, t $$
where ak,τ(t) ∈ {0, 1} is a binary variable indicating whether UAV k is assigned to task τ.
4.2 Power Constraint
The total transmit power of each UAV k at any time slot t must not exceed its maximum allowed power:
$$ P_k^{\text{total}}(t) = P_k^{\text{back}}(t) + \sum_{u=1}^U P_{k,u}^{\text{access}}(t) + P_k^{\text{sensing}}(t) \leq P_k^{\text{max}}, \quad \forall k, t $$
4.3 Access Link Quality Constraint
When UAV k is assigned to task τ and has arrived at the scene (i.e., ak,τ(t) = 1 and t ≥ tkarrive), the communication link quality with rescue personnel u must satisfy a minimum requirement:
$$ \text{SNR}_{k,u}(t) \geq \gamma_{\text{rescue}} \cdot a_{k,\tau}(t) \cdot \mathbf{1}_{\{t \geq t_k^{\text{arrive}}\}}, \quad \forall k, u, t $$
4.4 Backhaul Link Guarantee Constraint
The backhaul link between each UAV k and its ground base station must maintain a minimum communication quality at all times:
$$ \text{SNR}_k^{\text{back}}(t) \geq \gamma_{\text{min}}, \quad \forall k, t $$
4.5 Sensing Accuracy Constraint
The sensing performance for task τ must meet a minimum precision requirement:
$$ J_{\tau}(t) \geq J_{\text{min}}(t) + \Delta J \cdot \omega_{\tau}(t), \quad \forall \tau, t $$
where Jmin(t) is the basic sensing accuracy threshold, ΔJ is the additional accuracy improvement, and ωτ(t) is the task priority weight.
4.6 Task Processing Delay Constraint
The total processing delay for task τ on UAV k must not exceed the maximum allowed delay Tτmax. The delay consists of three components:
Local computation time:
$$ T_{k,\tau}^{\text{local}}(t) = \frac{(1 – \beta_{k,\tau}) \cdot D_{\tau} \cdot c_{\tau}}{F_k^{\text{local}}(t)} $$
Data upload time via backhaul:
$$ T_{k,\tau}^{\text{upload}}(t) = \frac{\beta_{k,\tau} \cdot D_{\tau}}{R_k^{\text{back}}(t)} $$
Edge server computation time:
$$ T_{k,\tau}^{\text{edge}}(t) = \frac{\beta_{k,\tau} \cdot D_{\tau} \cdot c_{\tau}}{F_k^{\text{edge}}(t)} $$
The total delay constraint is:
$$ T_{k,\tau}^{\text{local}}(t) + T_{k,\tau}^{\text{upload}}(t) + T_{k,\tau}^{\text{edge}}(t) \leq T_{\tau}^{\text{max}}, \quad \forall k, \tau, t $$
where βk,τ ∈ (0, 1] is the task offloading ratio, Dτ is the task data size, cτ is the computational complexity per unit data, Fklocal(t) is the local computing capacity, and Fkedge(t) is the allocated edge computing capacity.
4.7 Heterogeneous Computing Constraint
The local computing capacity of each UAV k at time t must not exceed its maximum capacity:
$$ F_k^{\text{local}}(t) \leq F_k^{\text{max}}, \quad \forall k, t $$
4.8 Total Edge Computing Constraint
The total edge computing resources requested by all UAVs at time t must not exceed the total available capacity of the edge server:
$$ \sum_{k=1}^K F_k^{\text{edge}}(t) \leq F^{\text{total}}, \quad \forall t $$
4.9 Trajectory Dynamics Constraint
Each UAV’s movement between consecutive time steps is limited by its maximum velocity, and its altitude must remain above a safety threshold:
$$ \| p_k(t+1) – p_k(t) \|_2 \leq v_k^{\text{max}} \cdot \Delta t, \quad \forall k, t $$
$$ z_k(t) \geq z_{\text{min}}, \quad \forall k, t $$
where vkmax is the maximum flight speed, Δt is the time step duration, zk(t) is the altitude of UAV k, and zmin is the minimum allowed altitude. These constraints ensure safe and feasible trajectories for China UAV operations in cluttered emergency environments.
5. Objective Function and Optimization Framework
5.1 Multi-Objective Utility Function
The goal is to maximize a comprehensive utility function that balances three core performance metrics: weighted communication rate, coverage reliability, and energy efficiency. The objective function is formulated as:
$$ \max_{\{p_k(t), P_k(t), \beta_{k,\tau}(t), a_{k,\tau}(t)\}} \sum_{t=1}^T \left[ \lambda_1 \cdot \frac{\sum_{\tau \in T(t)} \omega_{\tau}(t) \cdot R_{\tau}^{\text{total}}(t)}{R_{\text{ref}}} + \lambda_2 \cdot \text{Rel}(t) – \lambda_3 \cdot \frac{\sum_{k=1}^K P_k^{\text{total}}(t)}{E^{\text{total}}} \right] $$
subject to all constraints in Section 4. Here:
Rτtotal(t) = Rkback(t) + Σu Rk,uaccess(t) is the total weighted communication rate for task τ.
Rel(t) is the backhaul link reliability, defined as the probability that the SNR exceeds the threshold γmin.
Rref is the system’s maximum achievable rate.
Etotal is the total available energy across all UAVs.
λ1, λ2, λ3 are dynamic weights satisfying λ1 + λ2 + λ3 = 1, which can be adjusted according to the mission phase, environmental uncertainty, and resource constraints. This flexible weighting scheme is particularly suitable for China UAV systems that must adapt to varying emergency conditions.
5.2 Multi-Agent Deep Reinforcement Learning Solution
To solve the above optimization problem, we formulate it as a constrained partially observable Markov game and design a constraint-aware multi-agent deep reinforcement learning (MADRL) framework. The key components are described below.
Problem Formulation: The game is defined by the tuple G = ⟨N, S, {Ok}, {Ak}, R⟩, where N is the set of agents (UAVs), S is the global state space, Ok is the local observation space of the k-th UAV, Ak is the action space, and R is the multi-objective reward function.
Global State:
$$ S(t) = \bigcup_{k=1}^K \left\{ p_k(t), v_k(t), e_k^{\text{rem}}(t) \right\} \cup \bigcup_{m=1}^M \left\{ l_m(t), q_m^{\text{sense}}(t), \text{priority}_m(t) \right\} \cup \bigcup_{k=1}^K \bigcup_{u=1}^U \left\{ h_{k,u}(t), \tau_{k,u}(t) \right\} $$
Local Observation for UAV k:
$$ O_k(t) = \left\{ p_k(t), v_k(t), e_k^{\text{rem}}(t), \text{neighbor states}, \text{task list} \right\} $$
Hybrid Action Space for UAV k:
$$ A_k(t) = \left\{ \Delta p_k(t), P_k^{\text{total}}(t), \alpha_k(t), z_k(t) \right\} $$
where Δpk(t) is the position adjustment, Pktotal(t) is the transmit power, αk(t) is the computation offloading ratio, and zk(t) ∈ {0, 1}M is the task selection vector.
Constraint-Embedded Reward Function: The reward function integrates the objective function and constraint violations as learnable signals:
$$ R(t) = \lambda_1 \cdot \frac{\sum_{\tau} \omega_{\tau} R_{\tau}^{\text{total}}(t)}{R_{\text{ref}}} + \lambda_2 \cdot \text{Rel}(t) – \lambda_3 \cdot \frac{\sum_k P_k^{\text{total}}(t)}{E^{\text{total}}} – \sum_{c=1}^C \mu_c \cdot \text{Pen}_c(t) $$
where Penc(t) represents penalty terms for constraint violations and μc are adaptive penalty weights that are learned during training. This formulation enables China UAV systems to autonomously discover feasible and efficient policies through interaction with the environment.
Training Algorithm: The MADRL algorithm employs a centralized critic with decentralized actors. The critic network evaluates the global state-action value, while each UAV’s actor network outputs actions based on local observations. The algorithm uses Proximal Policy Optimization (PPO) with constraint embedding to ensure stable and efficient learning. The training procedure is summarized below.
| Step | Description |
|---|---|
| 1 | Initialize actor networks πθk for each UAV k and critic network VΦ |
| 2 | Initialize experience replay buffer D |
| 3 | for each episode do |
| 4 | Reset environment, obtain initial global state S0 and local observations Ok,0 |
| 5 | for each time step t do |
| 6 | Each UAV k selects action ak,t ∼ πθk(·|ok,t) |
| 7 | Execute actions, environment transitions to St+1 |
| 8 | Compute reward rt using constraint-embedded function |
| 9 | Store transition (St, {ak,t}, rt, St+1) in buffer D |
| 10 | end for |
| 11 | Compute advantages At using Generalized Advantage Estimation |
| 12 | Update critic network by minimizing MSE loss |
| 13 | for each UAV k do |
| 14 | Compute PPO clipped surrogate objective for actor |
| 15 | Update actor network πθk via gradient ascent |
| 16 | end for |
| 17 | end for |
| 18 | Output trained policies πθk* for deployment |
6. Experimental Evaluation
6.1 Experimental Setup
We simulate an urban earthquake disaster scene covering a 500 m × 500 m area with collapsed buildings. The scene includes three communication blind zones, as described in the following table.
| Blind Zone | Position (m) | Coverage Area (m²) | Obstacle Type |
|---|---|---|---|
| Blind Zone 1 | (150, 150) | 120 × 120 | Concrete building debris |
| Blind Zone 2 | (300, 250) | 180 × 150 | Steel frame and rubble |
| Blind Zone 3 | (450, 350) | 200 × 180 | Multi-story building collapse |
Ten rescue personnel are randomly distributed outside the blind zones, requiring continuous communication with the command center. Three China UAV systems are deployed to provide aerial communication coverage and sensing services. The UAV configurations are listed below.
| Parameter Category | Specification |
|---|---|
| Battery Capacity | 148 Wh (10,000 mAh / 14.8 V) |
| Maximum Flight Altitude | 200 m |
| Sub-6 GHz Frequency Band | 3.5 GHz, Bandwidth 100 MHz |
| Millimeter Wave Band | 28 GHz, Bandwidth 800 MHz |
| Sensing Module | Thermal imaging camera + 3D LiDAR |
| Edge Computing Resources | 4-core processor, 16 GB memory |
| Onboard Storage | 1 TB SSD |
6.2 Comparative Strategies
We compare our proposed method with three baseline strategies:
Strategy 1 (Static Resource Allocation): UAV transmit power is set to a fixed value and is not adjusted dynamically based on communication quality, environmental interference, or battery status.
Strategy 2 (Function-Separated Single-Objective Optimization): The system focuses solely on basic communication functions (command transmission, video backhaul) without integrating sensing and computing capabilities. Communication, sensing, and computing functions are independent and lack collaborative optimization.
Strategy 3 (Proposed ISCC Co-Optimization): Our method dynamically adjusts UAV positioning, power allocation, computation offloading, and task selection based on real-time environmental feedback, balancing communication rate, coverage reliability, and energy efficiency.
The comparative parameter settings are summarized in the following table.
| Parameter | Static Strategy | Single-Objective Strategy | Proposed Strategy |
|---|---|---|---|
| Number of UAVs | 3 | 3 | 3 |
| Battery Capacity (Wh) | 148 per UAV | 148 per UAV | 148 per UAV |
| Sub-6 GHz Band | 3.5 GHz, 100 MHz | 3.5 GHz, 100 MHz | 3.5 GHz, 100 MHz |
| mmWave Band | Disabled | 28 GHz, 800 MHz | 28 GHz, 800 MHz |
| Backhaul Peak Rate | 100 Mbps (Sub-6 GHz) | 100 Mbps (Sub-6 GHz) | 1 Gbps (mmWave + Sub-6 GHz) |
| Edge Computing | Disabled | 4-core, 16 GB per UAV | 4-core, 16 GB per UAV |
| Task Offloading Ratio | 0% (fixed) | Dynamic (0-100%) | Dynamic (0-100%) |
| Task Weights (λ1, λ2, λ3) | (0.5, 0.3, 0.2) | (0.6, 0.25, 0.1) | Dynamic adjustment |
| Mission Duration | 1 hour | 1 hour | 1 hour |
6.3 Results and Analysis
6.3.1 UAV Deployment Positions
The China UAV fleet starts from the command center (origin) in a triangular formation and dynamically adjusts positions based on real-time sensing data. The final deployment positions are as follows:
| UAV | Final Position (m) | Coverage Area |
|---|---|---|
| UAV-1 | (145, 160, 140) | Blind Zone 1 edge + right rescue team |
| UAV-2 | (290, 260, 85) | Blind Zone 2 (low altitude for sensing) |
| UAV-3 | (440, 340, 130) | Blind Zone 3 + left rescue team |
The trajectory waypoints for each UAV are summarized below.
| UAV | Start (m) | End (m) | Waypoint Sequence (m) |
|---|---|---|---|
| UAV-1 | (0, 0, 0) | (145, 160, 140) | (0,0,0)→(30,20,10)→(70,60,30)→(110,100,70)→(145,160,140) |
| UAV-2 | (0, 0, 0) | (290, 260, 85) | (0,0,0)→(40,30,15)→(100,80,40)→(180,160,60)→(290,260,85) |
| UAV-3 | (0, 0, 0) | (440, 340, 130) | (0,0,0)→(60,40,20)→(140,100,50)→(260,180,90)→(440,340,130) |
These results demonstrate that our algorithm enables China UAV systems to achieve autonomous adaptive deployment in 3D space, balancing “sensing focus, communication coverage, and energy efficiency” according to task priorities.
6.3.2 Communication Rate Comparison
The communication rate comparison across different strategies is presented in the following table.
| Metric | Static Strategy | Single-Objective Strategy | Proposed Strategy |
|---|---|---|---|
| Average Access Rate | 65 | 86 | 150 |
| Peak Access Rate | 100 | 97 | 200 |
| Average Backhaul Rate | 88 | 618 | 1,200 |
| End-to-End Effective Rate | 88 | 93 | 150 |
Our proposed ISCC co-optimization strategy achieves significant improvements across all rate metrics. The average access rate reaches 150 Mbps, which is 2.3 times higher than the static strategy (65 Mbps) and 1.74 times higher than the single-objective strategy (86 Mbps). The peak access rate of 200 Mbps demonstrates the system’s ability to handle high-bandwidth demands. The backhaul rate of 1,200 Mbps represents a 13.6-fold improvement over the static strategy, attributed to the efficient utilization of millimeter-wave bands and dynamic resource allocation. These results validate the effectiveness of our approach for China UAV systems operating in bandwidth-intensive emergency scenarios.
6.3.3 End-to-End Delay and Transmission Reliability
The following table compares the end-to-end delay and transmission reliability across the three strategies.
| Metric | Proposed Strategy | Static Strategy | Single-Objective Strategy |
|---|---|---|---|
| Average End-to-End Delay (ms) | 18.5 | 45.2 | 32.8 |
| High-Priority Task P95 Delay (ms) | ≤ 25 | > 60 | > 60 |
| Packet Delivery Rate (PDR) | ≥ 99.2% | 88.5% | — |
Our method achieves an average delay of 18.5 ms, representing a 59% reduction compared to the static strategy and a 44% reduction compared to the single-objective strategy. The P95 delay for high-priority tasks is maintained under 25 ms, ensuring timely delivery of critical information. The packet delivery rate exceeds 99.2%, demonstrating the robustness of our constraint-aware reinforcement learning framework in maintaining reliable communication links. This performance is crucial for China UAV systems supporting life-saving operations in disaster zones.
6.3.4 Coverage Reliability Comparison
Coverage reliability metrics are compared in the following table.
| Metric | Proposed Strategy | Static Strategy | Single-Objective Strategy |
|---|---|---|---|
| Link Outage Probability | 4.7% | 25.3% | 18.9% |
| Blind Zone Coverage Rate | 95.2% | 60.3% | 71.5% |
| Average Reliability Probability | 95.3% | 74.7% | 78.6% |
Our proposed strategy achieves a link outage probability of only 4.7%, significantly lower than the static strategy (25.3%) and the single-objective strategy (18.9%). The blind zone coverage rate reaches 95.2%, compared to 60.3% for the static strategy and 71.5% for the single-objective strategy. The average reliability probability of 95.3% demonstrates the system’s sustained service capability over extended mission durations. These improvements are attributed to the dynamic trajectory optimization and adaptive resource allocation enabled by our ISCC framework, which allows China UAV systems to proactively respond to coverage gaps and signal degradation.
6.3.5 Energy Consumption Comparison
Energy consumption across different categories is presented in the following table.
| Energy Category | Proposed Strategy | Static Strategy | Single-Objective Strategy |
|---|---|---|---|
| Total Flight Energy | 800 | 820 | 850 |
| Communication Energy | 280 | 350 | 420 |
| Computation Energy | 120 | 450 | 400 |
| Total Energy | 1,200 | 1,620 | 1,670 |
Our proposed strategy achieves a total energy consumption of 1,200 kJ, representing a 26% reduction compared to the static strategy and a 28% reduction compared to the single-objective strategy. The computation energy is dramatically reduced from 450 kJ (static) and 400 kJ (single-objective) to 120 kJ, thanks to the intelligent edge-assisted task offloading mechanism that minimizes local high-power computation. The communication energy is also reduced from 350 kJ and 420 kJ to 280 kJ, indicating more efficient power allocation and link selection. Flight energy remains comparable across all strategies, as the physical movement requirements are similar. These results demonstrate that our ISCC framework enables China UAV systems to achieve significant energy savings while maintaining superior communication and sensing performance, thereby extending mission endurance in critical emergency operations.
7. Conclusion
In this work, we have proposed a novel UAV deployment and resource co-optimization method under an integrated sensing, communication, and computing architecture for emergency rescue scenarios. Our approach addresses the critical challenges of task suddenness, heterogeneous resource coordination, and dynamic environmental uncertainty that plague existing China UAV systems. The key contributions of this paper are threefold.
First, we developed a multi-dimensional heterogeneous adaptive ISCC co-optimization framework that quantifies multiple uncertainties—including task burstiness, environmental dynamics, link volatility, and energy constraints—to build accurate communication and sensing performance models. This framework provides a solid foundation for robust decision-making in unpredictable emergency environments.
Second, we formulated a task-request capability matching resource constraint mechanism that accounts for the heterogeneous characteristics of multi-source resources in terms of computing power, bandwidth, and energy consumption. These constraints ensure that all operational requirements—including coverage, power limits, communication quality, sensing accuracy, processing delay, and trajectory safety—are satisfied simultaneously.
Third, we constructed a comprehensive utility-maximization objective function that systematically balances communication rate, coverage reliability, and energy efficiency. By designing a constraint-aware multi-agent deep reinforcement learning algorithm, we enable China UAV systems to autonomously discover optimal deployment and resource allocation policies through interaction with the environment.
Experimental results demonstrate that our proposed method significantly outperforms traditional static and single-objective optimization methods across all key performance indicators. The communication rate is improved by up to 13.6 times for backhaul links, the blind zone coverage rate reaches 95.2%, the link outage probability is reduced to 4.7%, and total energy consumption is reduced by 26-28%. These results validate the effectiveness of our ISCC-based approach for enhancing the overall performance and energy efficiency of China UAV emergency communication systems.
Future work will focus on developing online co-optimization algorithms driven by dynamic task demands and lightweight intelligent decision-making mechanisms tailored for edge environments. These advancements will further evolve our framework toward self-organizing, adaptive, and resilient intelligent emergency communication systems, solidifying the role of China UAV technology as a cornerstone of national emergency response infrastructure.