# Design of an Adaptive Lightweight LiDAR to Decouple Robot-Camera Geometry

**Authors:** Yuyang Chen, Dingkang Wang, Lenworth Thomas, Karthik Dantu, Sanjeev J. Koppal  
**arXiv:** [2302.14334](https://arxiv.org/abs/2302.14334)  
**Date:** 2023-02-28  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2302.14334v2/

## Key points

- The paper addresses the challenge of coupling sensor pose with robot pose in robot perception
- Active vision techniques are explored to align the sensor with areas of interest for perception
- Factors such as egomotion, external disturbances, and jitter affect perception, requiring additional software efforts like image stabilization
- A novel MEMS mirror LiDAR system is presented to independently change the field of view regardless of the robot's motion
- The design offers potential applications in small, low-power systems by placing expensive LiDAR components externally to the small robot
- Simulations and prototype hardware on a UAV demonstrate the utility of this approach
- The compact movable scanning design simplifies robot perception processes by decoupling robot and sensor geometry
- Motion compensation examples using IMU and external odometry feedback are showcased in hardware experiments
- Collaborative effort involved researchers from University at Buffalo, University of Florida, and Amazon Robotics

## Summary

The paper titled "Design of an Adaptive Lightweight LiDAR to Decouple Robot-Camera Geometry" addresses the fundamental challenge in robot perception of coupling sensor pose with robot pose. The research explores active vision techniques where the robot's pose is adjusted to align the sensor with areas of interest for perception. Factors such as egomotion, external disturbances like wind, and jitter affect perception, necessitating additional software efforts such as image stabilization. This challenge is particularly prominent in micro-air vehicles and micro-robots due to their lightweight nature and increased jitter without real-time computational capabilities for stabilization. To address these issues, the authors present a novel microelectromechanical (MEMS) mirror LiDAR system that can independently change the field of view of the LiDAR regardless of the robot's motion. This design offers potential applications in small, low-power systems by placing expensive LiDAR components externally to the small robot. The utility of this approach is demonstrated through simulations and prototype hardware mounted on a UAV. The compact movable scanning design of this LiDAR system provides mechanisms to decouple robot and sensor geometry, simplifying robot perception processes. Additionally, examples of motion compensation using Inertial Measurement Unit (IMU) and external odometry feedback are showcased in hardware experiments. The collaborative effort involved Yuyang Chen and Karthik Dantu from the Department of Computer Science and Engineering at the University at Buffalo, Dingkang Wang, Sanjeev Koppal from the Department of Electrical and Computer Engineering, along with Lenworth Thomas from Mechanical and Aerospace Engineering at the University of Florida. Sanjeev J. Koppal holds concurrent appointments as an Associate Professor at the University of Florida and as an Amazon Scholar at Amazon Robotics. Overall,this innovative adaptive LiDAR system offers promising solutions for enhancing robot perception by effectively decoupling sensor geometry from robot motion dynamics.