|

Loading ...
|
|
|
|
pages views since 05/19/2016 : 142970
· Members : 7
· News : 806
· Downloads : 0
· Links : 0
|
|
|
|
The Surprising Method for Creating Smarter Robots
|
|
|
Posted by Okachinepa on 05/02/2024 @


Courtesy of SynEvol
Credit: Northwester University
A novel artificial intelligence (AI) algorithm has been created by Northwestern University engineers especially for intelligent robotics. The new approach could greatly increase the usefulness and safety of robots for a variety of applications, such as self-driving cars, delivery drones, home assistants, and automation, by assisting them in learning complex skills quickly and reliably.
The approach known as Maximum Diffusion Reinforcement Learning (MaxDiff RL) is successful because it motivates robots to investigate their surroundings as haphazardly as possible in order to acquire a variety of experiences. The quality of the information that robots gather about their immediate environment is enhanced by this "designed randomness." Additionally, simulated robots showed faster and more effective learning by employing higher-quality data, which enhanced their overall performance and reliability.
Northwestern's new algorithm produced consistently better simulated robots than state-of-the-art models when evaluated against competing AI platforms. Because of how well the new algorithm functions, robots have been able to learn new tasks and do them successfully on their first try. This is in sharp contrast to existing AI models that allow for more gradual learning via trial and error.
The study's lead researcher, Thomas Berrueta of Northwestern, stated, "Other AI frameworks can be somewhat unreliable." They will occasionally succeed admirably at a task, but other times they may fall flat. With our framework, you can expect your robot to perform precisely as instructed each and every time it is turned on, provided that it is capable of completing the task at all. In a world where artificial intelligence is becoming more and more important, this facilitates the interpretation of robot achievements and failures.
Berrueta is a mechanical engineering Ph.D. candidate at the McCormick School of Engineering and a Presidential Fellow at Northwestern. The paper's senior author is robotics specialist Todd Murphey, a professor of mechanical engineering at McCormick and Berrueta's advisor. Co-authoring the paper with Berrueta and Murphey was Allison Pinosky, a Ph.D. candidate working in Murphey's lab.
Scientists, engineers, and researchers employ vast amounts of human-curated and filtered big data to train machine-learning algorithms. Through trial and error, AI gains knowledge from this training set and eventually achieves optimal outcomes. This method does not work for embodied AI systems, such as robots, but it does work effectively for disembodied systems, such as ChatGPT and Google Gemini (previously Bard). Instead, robots gather data on their own without the assistance of human curators.
According to Murphey, "traditional algorithms are incompatible with robotics in two distinct ways." "First, in a world where physical laws do not apply, disembodied systems can exploit this. Secondly, there are no repercussions for individual mistakes. The only thing that counts for computer science applications is that they work the majority of the time. One mistake in robotics could have disastrous consequences.
Berrueta, Murphey, and Pinosky set out to create a novel algorithm that guarantees robots will gather high-quality data while they are in motion in order to bridge this gap. Fundamentally, MaxDiff RL instructs robots to move with greater randomness in order to gather comprehensive, varied data about their surroundings. Robots learn through self-selected random experiences, gaining the essential abilities to do practical tasks.
The researchers tested the novel algorithm by contrasting it with the most advanced models available at the time. The researchers trained simulated robots to carry out a number of routine activities using computer simulations. Robots employing MaxDiff RL generally picked up new skills more quickly than those using other models. Additionally, they completed duties far more frequently and dependably than others.
Even more astounding, maybe: Robots employing the MaxDiff RL technique frequently completed a task accurately the first time. And that was even before they had any prior information.
"Our robots exhibited greater speed and agility, enabling them to proficiently apply their acquired knowledge to novel scenarios," stated Berrueta. "This is a huge benefit for real-world applications where robots cannot afford endless time for trial and error."
MaxDiff RL is a generic algorithm, which makes it suitable for many different applications. In the end, the researchers hope it will open the door for trustworthy decision-making in smart robotics by addressing fundamental problems impeding the industry.
Pinosky stated, "This doesn't have to be used only for robotic vehicles that move." Additionally, it might be applied to stationary robots, such a kitchen robotic arm that picks up the skill of loading the dishwasher. The role of embodiment becomes even more important to take into consideration during the learning process as tasks and physical settings are more intricate. This is a significant step toward the development of actual systems that can do more intricate and engaging activities.
|
|
|