CATS is a participant in the JASA AI Research and Technology Committee, in which it was chosen to head the NEDO project “Next-gen AI/Core Robot Technology Development/Next-gen AI Technology Fields/Realization of Embedded Technology Capable of Processing Ontological Reasoning in Real-time, and Their Application to Safe/Reliable Fields”.
We are testing ways to speed up the time it takes for processing from when a pedestrian walking across a cross-walk is detected by an self-driving car, to when a decision is made to stop the vehicle.
As cars can travel meters in the space of a second, self-driving cars are required to make extremely quick situational judgments.
Current self-driving systems apply a series of rules based on ontological reasoning to determine when to stop the car. While these decisions took several seconds to make, we were able to demonstrate that speed improvements could be expected by expressing decision rule semantics in a decision table and using this decision table to generate code.
Looking ahead, we will work toward establishing a technology to convert ontological reasoning to an applicable decision table.
These are teaching aids used for learning rule-based architecture through the designing of complex self-driving agents.
Machine learning and rule-based architecture are two main methods to designing agents. The concept behind rule-based architecture, with its clearly defined relationships between cause and effect, can be learned from to create seemingly complex self-driving agents.
In the example shown in the video, a self-driving agent is created from a set of 27 rules.
Here we verify the feasibility of a hybrid AI using a commercially available robot.
In terms of the method used by this hybrid AI to generate an agent by emphasizing the data driven AI and the knowledge-based AI, in this example, the results of signal ‘recognition’ (data driven AI) by the commercial robot are used to make a ‘decision’ on the next course of action (knowledge-based AI realized from rule based architecture).
This is used to determine the status of a project and issues present by using a graph database to inquire about information required, analyze relevancy and provide a visual representation of a project in the form of a graph using software development project data.
Graph data models are used in question and answer systems, and the increased flexibility in data searches made possible through graph pattern matching enables users to respond to a broad range of queries, such as “what tasks and development deliverables has this developer been involved in” and “who was involved in the development of features for which bugs occurred during a set period”.
Further, this is ideally suited to the visualization of inquiry results, which is greatly beneficial to understanding the configuration of project tasks and software architecture.