Service robots with integrated artificial intelligence usually receive training in simulated environments before being sent out into the real world. Unfortunately, AI can encounter unforeseen situations in the real world that weren’t covered in simulated environments, and these become blind spots in its decision making. To combat this problem, some service robots are receiving on-the-job training.
Service Robots Get Real-World Training
To develop machines that know what to do in real-world scenarios that weren’t anticipated in simulation, computer scientists are subjecting service robots to post-simulation programs where a human demonstrator helps the artificial intelligence system find gaps in its education. Engineers are then able to develop better simulations for future service robots.
The service robot enters a probationary period where it notes environmental factors that influence the observed human’s actions. Whenever the human trainer does something the AI software doesn’t expect, the AI scans the surroundings for previously unknown elements.
An example of this may be a human waiting to enter a crosswalk. Why would this human wait rather than proceeding? The person may have the right-of-way and have a clear path, and there may be no approaching vehicles. But, perhaps the human paused because sirens could be heard. The AI software can then associate the sound of a siren to keep it from crossing roads. It learns an important lesson from the human’s judgment.
Self-Aware Robots Go the Last 50 Feet
Autonomous vehicles could soon be navigating city streets to deliver parcels, groceries, and pizzas without a human in the driver’s seat, but those vehicles are confined to roadways. Engineers are working on the solution to this “curb-to-door problem.”
Soon, two-legged robots equipped with LiDAR may be walking up to and into buildings and homes (LiDAR is Light Detection and Ranging, which sends out laser beams to detect objects and avoid them). An autonomous vehicle could arrive at a destination, the customer service robot exits, grabs the package, and walks to the door. The autonomous vehicle can host a hefty suite of sensors and computing power for complex decision making.
This change can help service robots make the leap to self-aware robots. The self-aware robot already knows where it is before it’s activated at a destination, comparable to sitting on top of the vehicle observing all 360 degrees of its surroundings. If the customer service robot runs into trouble when making the delivery, it can communicate with the autonomous vehicle for help.
Why two-legs over wheels? Two-legged robots have advantages, like the capability of stepping over cracks, walking upstairs, and maneuvering more precisely. Testing and perfecting the service robot may give manufacturers an edge in the upcoming Robo-Taxi market.
Visit our Service Robots informational section to keep up to date with advances in technology and applications.