21.2 A 1.5μJ/Task Path-Planning Processor for 2D/3D Autonomous Navigation of Micro Robots

Autonomous micro robots have been deployed for various applications, ranging from unmanned package delivery to smart aerial surveillance. These robots possess intelligence for perception, make decisions based on the collected information, and take actions automatically [1], as illustrated in Fig. 21.2.1. Due to the limited battery lifetime of micro robots, energy-efficient processing for path planning is critical, especially in dynamic environments. In [2], a low-power path planning processor for 2D indoor navigation tasks through the iterative-deepening A * (IDA *) algorithm is proposed. IDA * is a grid-based planning algorithm, which finds a deterministic path on a pre-defined graph converted from a grid map. However, IDA * requires massive memory usage to store the entire map, leading to exponential complexity in memory storage and search time as the map size increases. For maps with an increased dimension and resolution, such as for navigation in large 3D space, IDA* usually fails to find a solution with an acceptable latency.