While we spend lots of dollars on designing new robots and autonomous systems, we need to be careful that we don't create machines that fail spectacularly when confronted with primitive, low-cost defeats (e.g., blinding optical sensors with paint bombs) and are unable to recover functionally without physical recovery and/or reboot.
Adversaries, especially asymmetric ones, are wily and adaptive and will look for the lowest-cost way to negate our advantage...often this is the small, simple thing the designers forget.Add Improvement
We need to design robotic and autonomous systems that fail gracefully and partially, that adapt while retaining a percentage of operational functionality even when confronted with novel countermeasures. Also, we need to red-team the heck out of any new system in every conceivable adversarial context to identify at least the low-hanging fruit of countermeasures and make our adversaries work to defeat our systems.
Advanced to the Marketplace.View this idea