Back to case studies
Embedded / Robotics
Jesture
Jesture enables operators to control rovers and remote vehicles using hand gestures captured by computer vision and mapped to safe motion primitives.
- Type
- Embedded / Robotics
- Year
- 2026
- Timeline
- 8 months
- Platforms
- Embedded, Web control
The problem
- Reliable hand-tracking in varied lighting and from different camera angles.
- Mapping noisy gesture signals to safe vehicle commands with deterministic fail-safes.
- Low-latency control loop for near-real-time responsiveness.
What we did
- Used a hybrid tracker combining neural pose estimation with lightweight Kalman smoothing.
- Designed a safety layer that converts high-level gestures into vetted motion primitives with automatic stop conditions.
- Deployed edge inference for low latency and a web dashboard for monitoring.
Where it landed
- Operators performed common manoeuvres faster with fewer control errors.
- The safety-first mapping made trials acceptable to compliance teams.
<100ms
Control latency (edge)
75%
Error reduction in trials
4
Supported vehicle classes
Jesture made remote control feel natural while keeping the system provably safe.
Hardware lead, robotics lab
