Low-cost autonomous robots suffer from limited onboard computing power, resulting in excessive computation time when navigating in cluttered environments. This paper presents Edge Accelerated Robot Navigation, or EARN for short, to achieve real-time collision avoidance by adopting hierarchical motion planning (HMP). In contrast to existing local or edge motion planning solutions that ignore the interdependency between low-level motion planning and high-level resource allocation, EARN adopts model predictive switching (MPS) that maximizes the expected switching gain w.r.t. robot states and actions under computation and communication resource constraints. As such, each robot can dynamically switch between a point-mass motion planner executed locally to guarantee safety (e.g., path-following) and a full-shape motion planner executed non-locally to guarantee efficiency (e.g., overtaking). The crux to EARN is a two-time scale integrated decision-planning algorithm based on bilevel mixed-integer optimization, and a fast conditional collision avoidance algorithm based on penalty dual decomposition. We validate the performance of EARN in indoor simulation, outdoor simulation, and real-world environments. Experiments show that EARN achieves significantly smaller navigation time and collision ratios than state-of-the-art navigation approaches.

翻译：低成本自主机器人因机载计算能力有限，在杂乱环境中导航时面临计算时间过长的问题。本文提出边缘加速机器人导航（简称EARN），通过采用分层运动规划（HMP）实现实时避障。与忽略低层运动规划与高层资源分配相互依赖性的现有局部或边缘运动规划方案不同，EARN采用模型预测切换（MPS）策略，在计算与通信资源约束下最大化相对于机器人状态与动作的期望切换收益。这使得每个机器人能够在本地执行的保证安全性（如路径跟踪）的点质量运动规划器与非本地执行的保证效率（如超车）的全形状运动规划器之间动态切换。EARN的核心在于基于双层混合整数优化的双时间尺度集成决策规划算法，以及基于惩罚对偶分解的快速条件避障算法。我们在室内仿真、室外仿真及真实环境中验证了EARN的性能。实验表明，与最先进的导航方法相比，EARN实现了显著更短的导航时间和更低的碰撞率。