The tsunami that struck the Fukushima Daiichi Nuclear Power Plant in March 2011 caused radioactive cesium to leak into the surrounding soil. Decontaminating this land will be a problematic legacy for generations to come — by some estimates, more than 20 million cubic meters of soil needs to be isolated. Researchers from Keio University are working on techniques to help with this clean-up and other dangerous situations by improving the physics behind systems for remote- and computer-controlled machinery1,2.
Moving mountains with a model
Disasters such as the Fukushima incident and volcanic eruptions can make environments too dangerous for construction workers to enter, prolonging repairs by weeks or months. Companies in Japan are spearheading the development of driverless equipment such as bulldozers, excavators, and trucks that work autonomously or are guided by operators stationed hundreds of kilometers away.
While advances in aerial drones and computer sensors are improving how these machines interact with the environment, their efficiency is poor compared with that of directly controlled equipment.
“Most of the work taking place in this area is devoted to manufacturing real unmanned machines,” notes Genya Ishigami at Keio University’s Department of Mechanical Engineering. “But we believe that research into high-fidelity simulators can solve some of the issues involved in planning an effective construction process.”
Ishigami cites an example of a typical challenge encountered by teleoperated excavators — remote operators cannot easily judge how much soil has been moved by the bucket, so may overload machines. Sensing how much material is in and around a bucket requires performing calculations in real time that are beyond the capabilities of most virtual control systems.
To produce an algorithm capable of quickly and reliably testing interactions between a bucket and the soil, the Keio team modified an existing model, which had originally been developed to simulate how robots walk on sand. Their technique calculates how much resistive force an object generates as it moves through granular particles, with data coming from parameters such as intrusion angle, penetration depth, and geometric factors that account for soil levels inside and outside the bucket during excavation.
Digging for answers
The team’s trials compared simulations of rotational and bulldozer-like excavator motions to the actions of a real bucket equipped with a grid of tiny sensors in a soil mechanics test bed. Measurements of the three-dimensional force distribution between the soil and bucket proved the validity of their model — the algorithm could predict the same trends in force experienced by a bucket during a half-minute excavation in just a fraction of a second.
“The sensor-embedded bucket directly measures the force distribution between the bucket and soil, which is unique,” says Ishigami. “It helps clarify the key issues we need to work on to improve our model’s accuracy.”
Because the team’s technique can be adapted to different shapes and sizes, it could be used in simulating other parts of construction machines, including wheels, tracks, and blades. Ishigami, with a background in robotic space exploration, also sees opportunities in extraterrestrial excavation and construction. “It would be interesting to apply the model on the Moon or Mars, where gravity is different,” he notes.