CSR: Medium: Attention-Driven Software Architecture for Integrated Perception and Planning in Autonomous Agents

虽然在开发自主系统方面取得了重大进展，但有效地处理和实时理解大量传感器数据仍然是一个挑战，特别是对于计算能力有限的小型机器人代理。为了解决这个问题，该项目将创建一个注意力驱动的软件架构，可以识别和优先处理来自传感器的关键信息，在考虑资源限制和环境不确定性的情况下及时做出决策。该体系结构将通过适应上下文和预测未来的行动来全面优化计算调度、感知和规划。本项目将取得的三个重要进展是：(1)自主计算管道的上下文自适应调度，(2)基于学习的感知来预测动态环境中的未来动作，以及(3)在存在不确定性的情况下基于预期动作的运动规划和决策。通过注意力驱动架构将这些组件相互关联，该项目将解决资源受限自主系统中与时间敏感场景相关的基本挑战。该项目将在各个领域产生更广泛的影响，包括智能农业、制造业、监控和机器人配送。提出的系统理论研究将使资源受限的机器人代理在快速变化的环境中具有安全、可靠和实时的自主能力。该项目计划建立教育项目，并与行业合作伙伴合作，为本科生提供研究机会。通过积极参与更广泛的社区和针对代表性不足的学生，这项研究将为未来的自主系统培养一支有才华和多样化的劳动力队伍。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

While significant progress has been made in developing autonomous systems, effectively processing and making sense of vast amounts of sensor data in real time remains a challenge, especially for small robotic agents with limited computing capabilities. To address this problem, this project will create an attention-driven software architecture that can identify and prioritize critical information from sensors, enabling timely decision-making while considering resource constraints and uncertainties in the environment. This architecture will holistically optimize computation scheduling, perception, and planning by adapting to the context and anticipate future actions. Three important advances to be made in this project are: (1) context adaptive scheduling of autonomous computation pipelines, (2) learning-based perception to anticipate future actions in dynamic environments, and (3) motion planning and decision making based on anticipated actions in the presence of uncertainty. By interrelating these components through the attention-driven architecture, this project will tackle fundamental challenges associated with time-sensitive scenarios in resource-constrained autonomous systems. This project will have broader impacts across various domains, including smart agriculture, manufacturing, surveillance, and robotic delivery. The proposed system-theoretic research will empower resource-constrained robotic agents with safe, reliable, and real-time autonomous capabilities in rapidly changing environments. The project plans to establish educational programs and engage with industry partners to provide undergraduate research opportunities. By actively involving the broader community and targeting underrepresented students, this research will cultivate a talented and diverse workforce for the future of autonomous systems.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.