Advanced robotics technology applied to mixed waste characterization, sorting and treatment

There are over one million cubic meters of radioactively contaminated hazardous waste, known as mixed waste, stored at Department of Energy facilities. Researchers at Lawrence Livermore National Laboratory (LLNL) are developing methods to safely and efficiently treat this type of waste. LLNL has automated and demonstrated a means of segregating items in a mixed waste stream. This capability incorporates robotics and automation with advanced multi-sensor information for autonomous and teleoperational handling of mixed waste items with previously unknown characteristics. The first phase of remote waste stream handling was item singulation; the ability to remove individual items of heterogeneous waste directly from a drum, box, bin, or pile. Once objects were singulated, additional multi-sensory information was used for object classification and segregation. In addition, autonomous and teleoperational surface cleaning and decontamination of homogeneous metals has been demonstrated in processing mixed waste streams. The LLNL waste stream demonstration includes advanced technology such as object classification algorithms, identification of various metal types using active and passive gamma scans and RF signatures, and improved teleoperational and autonomous grasping of waste objects. The workcell control program used an off-line programming system as a server to perform both simulation control as well as actual hardware control of the workcell. This paper will discuss the motivation for remote mixed waste stream handling, the overall workcell layout, sensor specifications, workcell supervisory control, 3D vision based automated grasp planning and object classification algorithms