Building Robots That Survive the Real World | Andy Diepen (Caladan)
Andy Diepen of Caladan Robotics describes how his team designs hardware and control systems to make robots reliable in demanding real-world environments at the Humanoids Summit.
In this talk
- 1:16
Robots share core hardware challenges
Diepen notes that robots range from industrial arms and warehouse vehicles to defense platforms and humanoids. Under the hood, they share electronics, control systems, sensors, actuators, and software. In his industrial automation work, cables moving back and forth around the clock became a major failure point.
- 6:30
Design for survival, not just function
Diepen says physical AI builds on established hardware, control loops, motors, and safety requirements. Caladan helps partners with mechanical and electrical design, embedded platforms, supply chains, and obsolescence planning. The goal is to make systems last in the environments where they will operate.
- 8:09
Prototype early and verify requirements
Rather than leave prototypes until late in development, Diepen’s team builds hardware early to expose failure modes and test it in the field. Its work includes requirements management, verification, and validation. Experience in aviation, automotive safety, and industrial safety informs how the team addresses standards and regulations.
- 9:47
Match engineering to the environment
Different settings create different demands: hospital robots need materials and joint grease suited to sanitizing and washdown, while mining equipment may stay dirty and still need to work. Caladan helps identify applicable regulations and address issues such as EMI and EMC. Diepen also stresses tracking software revisions and field changes so teams can learn from failures.
- 14:23
Build for demanding conditions
Diepen describes a commercial vehicle project that required meeting extensive OEM requirements, including reverse voltage tolerance and data bus specifications, as well as surviving full submersion. Another project delivered a 6,000-pound unmanned logistics vehicle for the U.S. Marines from a clean sheet to prototype in about 18 months. His team also built a jet engine data system that operated at 10,000 RPM, capturing 125 channels with wireless power and data.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.