Task-Oriented Data Compression for Multi-Agent Communications Over Bit-Budgeted Channels

Abstract

Various applications for inter-machine communications are on the rise. Whether it is for autonomous driving vehicles or the internet of everything, machines are more connected than ever to improve their performance in fulfilling a given task. While in traditional communications the goal has often been to reconstruct the underlying message, under the emerging task-oriented paradigm, the goal of communication is to enable the receiving end to make more informed decisions or more precise estimates/computations. Motivated by these recent developments, in this paper, we perform an indirect design of the communications in a multi-agent system (MAS) in which agents cooperate to maximize the averaged sum of discounted one-stage rewards of a collaborative task. Due to the bit-budgeted communications between the agents, each agent should efficiently represent its local observation and communicate an abstracted version of the observations to improve the collaborative task performance. We first show that this problem can be approximated as a form of data-quantization problem which we call task-oriented data compression (TODC). We then introduce the state-aggregation for information compression algorithm (SAIC) to solve the formulated TODC problem. It is shown that SAIC is able to achieve near-optimal performance in terms of the achieved sum of discounted rewards. The proposed algorithm is applied to a geometric consensus problem and its performance is compared with several benchmarks. Numerical experiments confirm the promise of this indirect design approach for task-oriented multiagent communications.

Centralized Control of a Multi-Agent System Via Distributed and Bit-Budgeted Communications

“…In [6], a similar problem is solved for distributed control and quantization, wherein, the authors disentangle the design of task-oriented communication policies and action policies given the aid of a hypothetical functional Π m * . In particular, the functional Π m * is a map from the vector space K c of all possible communication policies π c to the vector space K m of optimal corresponding control policy π m * (•).…”

“…(iii) effectiveness problem: given channel and network constraints, how accurately the communication symbols can help to fulfil the desired task? While the traditional communication design addresses the technical problem, recently, the semantic problem [2] as well as the effectiveness problem [4], [6]- [10] have attracted extensive research interest.…”

“…The focus of this work is on the effectiveness problem where in contrast to semantic level and technical level communication design, the performance of the communication system is ultimately measured in terms of the average return/cost linked to the task [7]. In the (task-)effectiveness problem, we are not concerned only about the communication of meaning but also about how the message exchange is helping the receiving end to improve its performance in the expected cost/reward of an estimation task [3], [9] or a control task [6]- [8], [10].…”

“…Thus, in control tasks the source of information -local observations of the agent -is often a stochastic process with memory -e.g. linear or Markov decision processes - [6], [7], [10]. In the estimation tasks, however, the source of information is often assumed to be an i.i.d.…”

“…Finally, numerical experiments are carried out on a geometric consensus task to evaluate the performance of the proposed schemes in terms of the optimality of the MAS's expected return in the task. ABSA-1 and ABSA-2 are compared with several other benchmark schemes introduced by [6], in a multi-agent scenario with local observability and 1 By an indirect algorithm here we mean an approach that is not dependent on our knowledge from a particular task. Indirect approaches are applicable to any/(wide range of) tasks.…”