Robot Collision Detection Acceleration with In-Memory Search Based on the Monolithic 3D Integration of 2D Transistors and Vertical RRAMs

Collision detection consumes over 90% of the total computation time in robot motion planning, making its optimization crucial. Previous hardware accelerators have struggled to meet the time requirement (<1 ms) in dynamic environment while supporting the data storage for large roadmaps. In this work, we design and fabricate the monolithic 3D integrated structure based on 2D transistors and vertical resistive random-access memories (VRRAMs) to enhance memory density, and accelerate collision detection based on in-memory search.