How to train your robot?
summary
Experiences from a robotics startup show concrete choices: using Gazebo for simulation to validate customer warehouses, training perception models, and trading off simulation fidelity against cost; doing exploratory testing across sensors, localization, navigation and fleet coordination; building automation scripts to ship logs and smoke-check component stacks that actually save hours; testing on a real 4.5-ton vehicle with safety prep; discovering surprising failure modes like sunbeams, relocated Wi‑Fi, slippery floors and a $3,000 pallet of toilet paper that break otherwise passing tests.
AI-written summary · about
chapters (17)
- 0:00Robotics testing journey
- 1:16High-risk environments importance
- 3:29Costs in embedded systems testing
- 5:00Product layers and robotics context
- 7:33Exploratory testing in embedded systems
- 10:04Unit and integration testing
- 10:47System-level testing with simulation
- 13:24Simulation development and real-time factors
- 14:30Validating business cases via simulation
- 15:33Cautious automation in robotics testing
- 16:35Testing autonomous forklifts
- 18:03Safety for vehicle testing
- 19:25Environmental factors in testing
- 21:57Load challenges in robotics
- 23:45Choosing automation in testing
- 26:16Key lessons from embedded systems
- 27:42Q&A on warehouse training
related talks
41mTesting Autonomous Mobile Robots
47m4 Elements of Successful Test Automation
43mWelcome to the Real World, Robot!
24mExploratory: Testing Humanely, Checking Robotically
46mDo-it-yourself testing lab
05m"Share of Experience : Test Automation"
21mCan robots test the sky?
24mTesting of IoT and related technologies