Abstract

Our current car‑centric transport systems are congested, polluting, expensive, and often fail on short trips and first/last‑mile access. Micro-mobility fills these gaps with smaller, cleaner, cheaper modes like bikes, e‑bikes, and scooters that work especially well for trips of a few miles or less. Over the past 5-6 years, we have built a perception stack to enable micro-mobility. Building robust perception systems for 24/7 operation is hard. We will outline how multi-sensory methods can provide the needed robustness. No single method is adequate to build a system. We use a mixture of traditional recognition methods and deep learning techniques to achieve the required performance. We will present our strategy for localization, tracking of road-users, and robust interaction in dynamic settings using a number of hybrid pattern recognition methods.

Biography

Henrik I Christensen is a distinguished professor of computer science and the director of robotics at UC San Diego. He has held faculty positions in Aalborg, Stockholm, and Atlanta prior to San Diego. He is the co-founder of six companies and serves as an advisor to agencies and companies across four continents. He is a fellow of IEEE and AAAS.