Adam's CV
Basics
| Name | Adam Imdieke |
| Label | Ph.D. Student in Computer Science |
| imdie022@umn.edu | |
| Phone | 507.321.3309 |
| Url | https://adamimd.github.io/ |
| Summary | Multisensory robot manipulation learning, focusing on novel hardware for robot perception and neural network architectures for manipulation policies. Current work includes developing a tactile skin for arms on high DoF robots (Boston Dynamics Spot arm) and Contact aware Inverse Kinematics to enhance whole-body environmental interaction. |
Education
Publications
-
2025.08.01 AugInsert: Learning Robust Visual-Force Policies via Data Augmentation for Object Assembly Tasks
IROS 2025
Co-Author. Leverages Force/Torque data, proprioception, and vision to learn robust insertion policies.
-
2025.04.29 SPARK-Remote: A Cost-Effective System for Remote Bimanual Robot Teleoperation
ICRA Workshop: Human-Centric Multilateral Teleoperation
Lead Author. Proposes haptic feedback and torque limiting controllers for our dual-arm UR5e robot arm to improve depth perception and bimanual manipulation loop closure.
-
2024.07.01 Talk Through It: End User Directed Manipulation Learning using feedback to Guide Robot Skill Acquisition
RA-L 2024
Co-Author. End User Directed Manipulation Learning using feedback to Guide Robot Skill Acquisition.
Skills
| Policy Learning | |
| NN based manipulation policies | |
| Diffusion learning | |
| Reinforcement learning | |
| Multisensory perception |
| Technical Skills | |
| Python | |
| C++ | |
| PyTorch | |
| Jax | |
| ROS/ROS2 | |
| Git | |
| Linux (~10 years) | |
| Network programming |
| Robot Control | |
| Inverse Kinematics | |
| Motion planning | |
| Force/Torque response | |
| Impedance control |
| Hardware Development | |
| CAD modeling | |
| 3D printing (SLA, FDM) | |
| PCB design | |
| SMD soldering | |
| Embedded systems |
| Simulation | |
| Mujoco | |
| PyBullet | |
| ROS | |
| Isaac Lab |
Projects
- 2025.05 - Present
Tactile Skin for Spot
A Novel, Low-cost 3D printed tactile skin for robotic arms to enhance whole-body environmental contact sensing.
- Leverages high Degree of Freedom robots to condition Inverse Kinematics null spaces to satisfy contact constraints
- Hardware development for real-time Contact aware Inverse Kinematics (ContactIK) enabling contact avoidance contact embracing behaviors
- 2025.09 - Present
Generative Models
Investigating Video Diffusion Models for Zero-Shot Robotic Manipulation Policy Learning
- Research project exploring generative AI for robotics
- 2023.11 - Present
Spot Natural Language Interface
Integrating LLM control of Boston Dynamics Spot, enabling natural language commands for long-horizon tasks.
- Novel human following capabilities, robust to dynamic environments and crowds
Interests
| Machine Learning | |
| Generative models | |
| Perception Models | |
| Manipulation Policy Learning | |
| Transformers | |
| Neural Architectures | |
| Imitation Learning (IL) | |
| Reinforcement Learning (RL) |
| Control | |
| Optimization-based control (IK, MPC) | |
| Modern Control | |
| System ID | |
| Filtering |
| Teleoperation | |
| Human-Robot Interaction | |
| Haptic Feedback | |
| VR Interfaces | |
| Low-latency Systems |