Issue link: https://iconnect007.uberflip.com/i/1228683
APRIL 2020 I SMT007 MAGAZINE 91 A simulation system invented at MIT to train driverless cars creates a photorealistic world with infinite steering possibilities, helping the cars learn to navigate a host of worst-case scenarios before cruising down real streets. Control systems for autonomous vehicles largely rely on real-world datasets of driving trajectories from human drivers. From these data, they learn how to emulate safe steering controls in a variety of situations. But real-world data from hazardous "edge cases," such as nearly crash- ing or being forced off the road or into other lanes, are— fortunately—rare. Computer programs, or "simulation engines," aim to imitate these situations by rendering detailed virtual roads to help train the controllers to recover. But the learned control from simula- tion has never been shown to transfer to reality on a full-scale vehicle. The MIT researchers tackle the problem with their photorealistic simulator, called Virtual Image Synthesis and Transformation for Autonomy (VISTA). The controller is rewarded for the distance it travels without crashing, so it must learn by itself how to reach a destina- tion safely, including regaining control after swerving between lanes or recovering from near-crashes. In tests, a controller trained within the VISTA simulator safely was able to safely deploy onto a full-scale driverless car and to navigate through previously unseen streets. In positioning the car at off-road orientations that mimicked various near-crash situations, the controller was also able to successfully recover the car back into a safe driv- ing trajectory within a few seconds. A paper describing the system has been published in IEEE Robotics and Automation Letters and will be presented at the upcoming ICRA conference in May. (Source: MIT News Office) System Trains Driverless Cars in Simulation Before They Hit the Road Integrate AI to Gemba Walks AI can improve the effectiveness of gemba walks. This can be done by evaluating real- time process data before gemba walks start. The objective is for manufacturing organiza- tions to design and implement an effective AI configuration in order to gather the right pro- cess data. AI can monitor and adjust to many more variables and data points faster objec- tively and more consistently than people can. The more AI learns, the more it can act and predict outcomes to subsequent changes. This rapid iteration allows people to see how some- thing in production can be improved, forecast the effects of change, and identify where there are bottlenecks or resource constraints. What would an enhanced gemba walk look like with AI? Table 2 shows a sample gemba checklist that integrates data evaluation from AI. With the ease of data gathering, thanks to AI, manufacturing organizations can realize more effective continual improvement efforts and specifically have successful gemba walks wherein managers and operators can eval- uate real-time data and make the necessary improvements to manufacturing process steps and work tasks accordingly. SMT007 Reference 1. "Gemba Walk: Where the Real Work Happens," Kan- banize. Alfred Macha is the president of AMT Partners. He can be reached at Alfred@amt-partners.com. To read past columns or contact Macha, click here.