Abhiraj Singh
Robotics Researcher & Builder
Building intelligent systems that enhance human capabilities.
Recent Projects
Research and engineering work in robotics and AI.
Sim-to-Real MPC Optimization for WalkOn F1: Lessons from Cybathlon 2024
Optimizing Model Predictive Control (MPC) walking trajectories for the WalkOn F1 exoskeleton that won Cybathlon 2024. An account of real-world sim-to-real challenges and the iterative solutions that bridged the gap between simulation and hardware.
Model Predictive Control for Powered Lower-Body Exoskeleton Robots
Advanced control methodology for powered lower-body exoskeletons using Model Predictive Control (MPC), enabling robust and adaptive operation in dynamic environments.
Identification of Physical Parameters of Body Segments
Estimating the center of mass (CoM) of the leg using the H-10 Hip Assistive Robot and comparing experimental and anthropometric methods for human-assistive robotics.
Trajectory Analysis & Optimization UI for Bipedal Robots (Real-Time Kinematics & Visualization)
A powerful interface for analyzing and optimizing bipedal robot motion trajectories in real-time. Integrates forward/inverse kinematics with visualization for processing millions of data points per minute, enabling rapid trajectory testing and optimization.
Writing
Thoughts on robotics, AI, and building technology.
Making a Robot Dog Stand: The Hidden Challenges Before Training Begins
Everyone shows the walking demos, but nobody talks about the hours spent getting a custom quadruped to simply stand upright. Here's what it takes to prepare a 6.65 kg robot dog for reinforcement learning.
The Model Context Protocol: Transforming AI Agents from Isolated Systems to Connected Ecosystems
An in-depth exploration of the Model Context Protocol (MCP) - the open standard that's enabling AI agents to break free from context limitations and interact dynamically with external systems.
Human-Centric Depth Vision: Intelligent Environment Recognition for Powered Exoskeletons
Exploring how depth cameras and IMU sensors can create intelligent terrain recognition systems that augment human decision-making in powered exoskeleton control, while maintaining the pilot as the ultimate authority.