Status: Planned. This track begins after the environmental controller, core reliability models, and deployment workflow are stable.
Why this belongs in the Lab
Robotics combines the same systems already being developed—embedded hardware, networking, sensing, cloud software, controls, reliability, predictive maintenance, and digital twins—inside a moving physical platform.
Primary toolchain
| Tool | Purpose |
|---|---|
| NVIDIA Isaac Sim | Physics-based robot and sensor simulation, manipulators, wheeled robots, perception testing, and synthetic environments. |
| Isaac Lab | Robot learning, reinforcement learning, policy training, and scalable simulation experiments. |
| ROS 2 | Robot middleware, messaging, control integration, hardware interfaces, and deployment. |
| Gazebo | Accessible robotics simulation and ROS 2 integration for early projects and comparison. |
| Python/C++ | Control logic, perception pipelines, experiment automation, testing, and telemetry analysis. |
Project sequence
- Import and control a wheeled robot in simulation.
- Model simulated cameras, depth sensing, IMU, lidar, and environmental sensors.
- Build autonomous navigation and obstacle-avoidance tests.
- Create a mobile environmental-inspection robot linked to the existing monitoring platform.
- Inject wheel, motor, battery, network, sensor, and localization failures.
- Build fault detection, isolation, recovery, and predictive-maintenance logic.
- Simulate robotic manipulation and inspection tasks.
- Generate synthetic data for perception and anomaly-detection models.
- Move selected software from simulation to a physical robot using ROS 2.
Evidence standard
Each study will publish the robot model, environment, assumptions, controller configuration, seeded failures, metrics, repeatable scripts, results, limitations, and the physical validation plan.
Questions this track should answer
- Can the robot complete its mission under sensor, actuator, power, or network degradation?
- Which failures can be detected, isolated, and recovered automatically?
- How closely does simulated sensor behavior match physical measurements?
- When should maintenance occur before performance becomes mission-limiting?
- What transfers successfully from simulation to hardware, and what does not?