End-to-end Neuromorphic Sensorimotor Control with Supervised Gait Learning for Autonomous Robot Navigation
Miniature robots are useful during disaster response and accessing remote or unsafe areas. They need to navigate uneven terrains without supervision and under severe resource constraints such as limited compute, storage and power budget. Biological locomotion relies on a hierarchical structure where Central Pattern Generators (CPGs) provide rhythmic primitives that are modulated by descending signals to produce complex motor patterns. Event-based sensorimotor control in edge robotics has potential to enable fully autonomous and adaptive robot navigation systems capable of responding to environmental fluctuations by learning new types of motion and real-time decision making to avoid obstacles.
In this talk, I will present my team’s work on a novel bio-inspired framework with a hierarchical control system, utilizing a tunable multi-layer neural network with a hardware-friendly CPG as the core coordinator to govern the precise timing of periodic motion. Autonomous operation is managed by a Dynamic State Machine (DSM) at the top of the hierarchy, providing the necessary adaptability to handle environmental challenges such as obstacles or uneven terrain. The multi-layer neural network uses a nonlinear neuron model which employs mixed feedback at multiple timescales to produce rhythmic patterns of bursting events to control the motors. The system was verified virtually in Gazebo simulation and validated in the real world via deployment on a Raspberry Pi platform. My team will demonstrate the proposed framework on the Petoi robot, which can autonomously learn walk and crawl gaits using supervised Spike-Time Dependent Plasticity (STDP) and surrogate gradient learning algorithms, transition between the learned gaits stored as new states, through the DSM for real-time gesture controlled navigation and obstacle avoidance. I will conclude with experimental results that demonstrate real-time neuromorphic control in low-power robotic systems.
Date and Time
Location
Hosts
Registration
-
Add Event to Calendar
Loading virtual attendance info...
- 3700 O'Hara Street
- Pittsburgh, Pennsylvania
- United States 15213
- Building: Benedum Hall
- Room Number: 102
- Starts 20 May 2026 04:00 AM UTC
- Ends 20 July 2026 12:00 AM UTC
- No Admission Charge
- Menu: Regular (no dietary restrictions), Vegetarian
Speakers
Rajkumar Kubendran
Rajkumar Kubendran is currently an Assistant Professor at the University of Pittsburgh, Department of Electrical and Computer Engineering with a secondary appointment in Bioengineering. He received his B.Tech from NIT Trichy in 2008 and M.S. degree from Purdue University in 2012. He received his Ph.D. degree from the University of California San Diego in 2020. His primary research focus is on energy-efficient Neuromorphic Computing Systems, spanning from devices to applications. His academic interests include low power analog and mixed signal circuit design with emerging memory devices to build event-driven architectures for computer vision and machine learning applications. He has demonstrated prototypes of dynamic vision sensors (DVS) and in-memory compute architectures with some of the best energy-efficiency metrics reported in literature. He has interned with multiple analog and RF design teams in industry, including Intel, IMEC Belgium, MaxLinear and Qualcomm. He is an IEEE Senior Member since 2023. Currently, his team is working on retina inspired cameras, custom AI hardware for neuromorphic computing, bio-inspired robots and programmable stimulators for biomedical applications.
Agenda
6:30 PM Gather and meal
7:00 PM Presentation
8:30 Adjourn