How about teaching social norms to robots? It seems like a different idea, but now it is possible. Palm Garden AI has created software called Coherence Guard. This is for service robots. The main feature of the software is to enhance the decision-making of robots when they interact with people. That means the software will make the robot able to understand the situation and respond accordingly. The best thing is that the software does not require a robot to change its existing system. It works well with it and checks whether an action is socially appropriate or not before performing it.

Humanoid robots are becoming more common day by day. From industries, schools, and hospitals to homes and offices, they are everywhere. This increase in their demand needs one more thing, and that is the ability to interact with humans in an appropriate way. The need is to make those robots understand how their actions can affect the people around them. That means when a robot interacts with a human being, it must know how the person feels toward the robot’s actions. Whether the person is comfortable or needs the robot to stop interaction or guidance, whatever task it is given to perform.

Palm Garden CEO Joachim Scheuerer said that the purpose behind Coherence Guard is not to replace motion planning. Neither is it intended to make any changes to a robot’s control system. It is only meant to review the robot’s planned actions. It is made to consider factors like timing, emotional signals, and personal space to analyze whether the action fits the situation or not. The purpose behind it is clear: to guide the robot to choose behavior that is appropriate for a real-life setting.

The software is built on the company’s ANATTA 9 behavior infrastructure. And it is designed to work with existing robot platforms, including ROS 2 and SDK/API-based systems. Palm Garden said that there is no need to replace the existing software. The technology works well with manufacturers’ existing systems.

The company has developed another system that does more than just understand objects and movements. This is called the Relational Infrastructure Framework (RIF). This system helps robots understand human roles and intentions. It also considers possible risks and effects that the robots’ actions may have on people. 

For example, if a person wants some space or they are not feeling comfortable with a robot, the system can stop the robot from further interaction. It can also guide the robot to step back completely or limit its interaction. This all depends on the situation.

According to the CEO, they got this idea from observing that most robots do very well when it comes to performing tasks. But they struggle while interacting with humans. The company studied these real-life situations for years and identified common behaviors that improve interactions. These include greeting people, asking for help, explaining various things to them, and, most importantly, staying away when someone is not interested in taking help or feels uncomfortable.

Palm Garden AI said that its purpose is not to replace existing systems but to support them. While hardware and collision-avoidance systems are responsible for physical security, Coherence Guard mainly targets the social interactions of a robot. It helps the robot decide when to pause or continue, or even ask for confirmation.

What are the recommendations by the company? Well, the company said that if you want to avoid delays in robot decision-making, you must install the software locally. However, you can use cloud services for updates and analytics.

Palm Garden is also working on introducing a commercial version of the software. This will be available for licensing. This will also have optional cloud-based services. The core technology is protected by a patent application. The company wants to ensure that its software works with a variety of robot platforms.

The company is also in discussion with other robotics companies. The most prominent ones are Robotera and Hanson Robotics. The company wants to test its software through simulations and pilot projects. Once done, the next step would be completing technical reviews. The company also has plans to expand its simulation testing and make the software available for industrial usage. Initially, they will only perform limited real-world tests to make sure that everything is going in the right direction.

One of the interesting things is that the company has no plans to make another robot. They said that the company only wants to add a better decision-making layer to robots. So that when they interact publicly, they adopt behavior that is socially more appropriate.