Adaptive Tuning of Multi-Robot Systems for The Effective Mapping and Tracking of Dynamic Environments

多机器人系统（MRS）将积极颠覆和促进几个关键经济部门，同时也为人类面临的一些重大挑战提供独特的解决方案，例如应对气候变化和提高农业生产力。然而，即使是MRS技术中的最新技术，与在真实世界动态设置中自主操作所需的速度相比，也遭受集体响应的缓慢速度。拟议的发现研究计划将通过利用机器人群体在面对不断变化的环境时所表现出的新颖的超响应集体行为的独特优势来解决这一关键问题。 拟议研究的长期目标是设计，开发和展示MRS在大规模，非结构化和快速发展的环境中运行的创新超响应集体行为。该研究计划将专注于四个短期目标：（1）利用异构群集进行高度响应的集体映射和跟踪;（2）通过自适应集体决策来优化平衡探索和利用;（3）通过多智能体对抗反向强化学习来推断新的快速集体机动;（4）通过多智能体对抗性反向强化学习来推断新的快速集体机动。（4）开发一类基准问题，以量化MRS操作在高度动态环境中的有效性。科学的方法包括理论，多智能体模拟和MRS实验的混合适合于满足我们的目标，如果要朝着我们的长期目标取得真实的进展，这是必不可少的。这项工作将分为两篇博士论文和两篇硕士论文，为HQP提供机会参与令人兴奋的，具有高社会影响力的前沿研究，直接造福加拿大的高科技产业。一些国际合作（美国，法国、新加坡、荷兰）将扩大和加强这一方案的成就。 拟议的Discovery研究计划及其对MRS超响应集体动力学的创新，预计将对现场机器人，农业和工业机器人的应用产生重大影响。最终，该计划的成功有可能成为快速新兴的多机器人自动化市场的催化剂，由于加拿大经济的几个关键部门的需求非常高，该市场正在经历急剧增长。这项研究计划所产生的创新可以为加拿大农业部门向超精密和自主农业的关键转型提供新的动力。为了充分利用机器人和自动化的潜力，实现农业5.0，需要在低技术准备水平上进行更多的基础研究。这正是这个发现研究计划的目的，特别是通过学习和适应，为异构系统的超快速协调提供新颖的进步和创新的解决方案。

Multi-robot systems (MRS) are about to positively disrupt and boost several key economic sectors, while also offering unique solutions to some of the grand challenges faced by humanity, such as dealing with climate change and increasing agricultural productivity. However, even the state-of-the-art in MRS technology suffers from a slow pace of collective response compared to what would be required to autonomously operate in real-world dynamic settings. The proposed Discovery research program will address this critical issue by exploiting the unique advantages of novel ultra-responsive collective behaviors exhibited by robot swarms when confronted with changing circumstances. The long-term aim of the proposed research is to design, develop and demonstrate innovative ultra-responsive collective behaviors of MRS operating in large-scale, unstructured, and fast-evolving environments. The research program will focus on four short-term objectives: (1) Leverage heterogeneous swarming for highly responsive collective mapping and tracking; (2) Optimally balance exploration and exploitation by means of adaptive collective decision-making; (3) Infer new swift collective maneuvers through multi-agent adversarial inverse reinforcement learning; (4) Develop a class of benchmark problems to quantify the effectiveness of MRS operations in highly dynamic environments. The scientific approaches consist of a blend of theory, multi-agent simulations, and MRS experiments appropriate to meet our objectives, and essential if real advances toward our long-term aim are to be made. The work will be partitioned between two doctoral and two master theses, providing opportunities for HQP to participate in exciting, cutting-edge research of high societal impact with direct benefit for Canada's high-tech industry. Several international collaborations (U.S., France, Singapore, The Netherlands) will amplify and augment the achievements of this program. The proposed Discovery research program with its innovations on ultra-responsive collective dynamics of MRS is expected to have a significant impact on applications in field robotics, agricultural and industrial robotics. Ultimately, the success of this program has the potential to be a catalyst in the fast-emerging market of multi-robot automation, which is experiencing a dramatic surge owing to a very high demand by several key sectors of Canada's economy. The innovations arising from this research program can give a fresh impetus to the pivotal transformation of Canada's agricultural sector toward ultra-precision and autonomous farming. To leverage the full potential of robotics and automation towards Agriculture 5.0, more fundamental research at low technological readiness levels is required. This is precisely what this Discovery research program is about, specifically novel advances and innovative solutions for the ultra-rapid coordination of heterogeneous systems, with learning and adaptation.