Computation and Communication Co-Design for Real-Time Monitoring and Control in Multi-Agent Systems

TL;DR

This paper tackles real-time monitoring/control over networks by jointly optimizing local computation (processing times $\tau_i$) and communication scheduling under an Age of Information (AoI) framework. It develops a scalable approach using a Lagrangian relaxation to decouple per-agent problems and derive threshold-based processing/scheduling decisions, followed by a Whittle-index policy to efficiently schedule transmissions in a multi-agent RMAB setting. The contributions include a general co-design framework, low-complexity algorithms with guaranteed structure, and validated improvements in two practical domains: multi-agent occupancy grid mapping and distributed ride sharing, with performance gains up to 82%. The results highlight the practical impact of AoI-driven co-design for real-time control and monitoring in large-scale networked systems, and point to future work on learning-based cost models and cross-agent coupling.

Abstract

We investigate the problem of co-designing computation and communication in a multi-agent system (e.g. a sensor network or a multi-robot team). We consider the realistic setting where each agent acquires sensor data and is capable of local processing before sending updates to a base station, which is in charge of making decisions or monitoring phenomena of interest in real time. Longer processing at an agent leads to more informative updates but also larger delays, giving rise to a delay-accuracy-tradeoff in choosing the right amount of local processing at each agent. We assume that the available communication resources are limited due to interference, bandwidth, and power constraints. Thus, a scheduling policy needs to be designed to suitably share the communication channel among the agents. To that end, we develop a general formulation to jointly optimize the local processing at the agents and the scheduling of transmissions. Our novel formulation leverages the notion of Age of Information to quantify the freshness of data and capture the delays caused by computation and communication. We develop efficient resource allocation algorithms using the Whittle index approach and demonstrate our proposed algorithms in two practical applications: multi-agent occupancy grid mapping in time-varying environments, and ride sharing in autonomous vehicle networks. Our experiments show that the proposed co-design approach leads to a substantial performance improvement (18-82% in our tests).

Figures (9)

• Figure 1: Example: four drones monitor different regions and send updates to a base station over a wireless channel. Each agent spends time $\tau_i$ processing the collected measurements before sending. A scheduling algorithm prioritizes transmissions to the base station. This paper focuses on the co-design of the processing times $\tau_i$ and the scheduling policy.
• Figure 2: AoI evolution for agent $i$. The agent acquires and processes new samples every $\tau_i$ time-slots. When the base station (B.S.) requests a new update, the agent sends the most recent sample that has finished processing, taking $r_i(\tau_i)$ time-slots for transmission. The variable $\delta_i^{(k)}$ represents the waiting time in the buffer for update $k$. Upon a new update delivery, the AoI at the base station $A_i(t)$ drops to the age of the delivered update.
• Figure 3: Multi-agent mapping over 9 regions: each agent monitors and builds a local grid map of a region, and sends map updates to a base station. The occupancy in the regions is time-varying. A scheduling policy specifies how to share the communication channel among the agents. Processing times specify how much time each agent spends in generating new map updates.
• Figure 4: Transition probabilities and optimal processing time allocations plotted for each region. The probabilities are plotted on a logarithmic scale while the processing times are plotted in number of time-slots.
• Figure 5: Performance of different scheduling policies vs. processing times $\tau$. Solid lines represent performance of different classes of scheduling policies as the processing time $\tau$ varies. The dotted lines represent the scheduling performance with processing times computed using Algorithm \ref{['alg:optimal-processing']}.

Theorems & Definitions (4)

• Remark 1: Task-related cost function
• Theorem 1
• Lemma 1
• Definition 1