How researchers timed a robot on a real battery pack
University of Birmingham and Faraday Institution researchers bought a real 2011 Nissan Leaf battery pack, 192 cells built into 48 modules, to test 2 robots taking it apart. Both were guided by a human operator throughout, not left to work alone. One was a single robot arm steered through a hand held controller that let the operator feel resistance as the robot searched for a bolt. The other used 2 identical robot arms that each matched a movement made by the operator, arm for arm. Stripping the same 4 modules took 25.8 minutes with the single arm and 14.7 minutes with the 2 matched arms, a cut in time of 43%, the number the researchers themselves report.
Both robots were guided by a human operator throughout on a real 2011 Nissan Leaf pack, not left to work alone. Researchers report a 43% cut in time using 2 matched arms instead of 1.
Show the numbers
| 1 arm, hand controller | 25.8 |
| 2 matched arms | 14.7 |
Where the robot still needs a person watching
The 2 arm robot was also tested for success, not only speed, across 5 tasks of rising difficulty, run many times on the same pack. It lifted the cover away every time, a 100% success rate. It cut a cable successfully 80% of the time and loosened 4 fasteners 85% of the time. It sorted a freed module into a bin only 60% of the time. On the hardest task, removing 8 bolts, it succeeded only 50% of the time, half of every attempt. The harder the task, the more often a human still had to finish it by hand.
Figures are for the 2 matched arm platform, across repeated trials on 1 real pack, not an industry wide figure. Removing more fasteners is the harder, less reliable task.
Show the numbers
| Lifting the cover away | 100 |
| Cutting a cable | 80 |
| Loosening 4 fasteners | 85 |
| Sorting a freed module | 60 |
| Removing 8 bolts | 50 |
AI is quite an important element during the process of disassembly because we have diversity in design of batteries.
Dr Alireza Rastegarpanah, research scientist in robotics and artificial intelligence, University of Birmingham. Source 3.
Why no single fix works for every pack
A welded seam that keeps a pack sealed while a car is built becomes the reason a recycler cannot get back inside without cutting, Rastegarpanah said in a separate interview. He added that a camera built to spot a fastener often cannot, because the metal is shiny, sits only partly in view, and shares a space built for putting a pack together, not for a robot to look inside. Dr Christoph Neef of the Fraunhofer Institute for Systems and Innovation Research coordinated a study finding that the European Union will need to recycle around 230 kilotons of battery material a year by 2030, rising to about 1,500 kilotons a year by 2040.
The boldest claim, and its limits
1 newer system goes further than any other in this file, claiming full autonomy with no human step at all. A dual arm robot took apart a 21 cell lithium ion pack, the size of a power tool or a small electric bicycle battery, not a full sized car pack, sensing where each cell sat on its own instead of being told in advance. It finished the whole job successfully in 8 of 10 attempts, in a mean of 6.0 minutes each, using a camera on its own wrist to correct its first guess as it moved. The result is not yet peer reviewed, and the pack tested is far smaller than a full sized car pack.
Oak Ridge National Laboratory, a United States government laboratory, makes a different claim, that its own line already removes a human from full sized pack disassembly once a robot first picks a pack up. In the time it takes to strip 12 battery stacks by hand, the laboratory says its automated line can handle 100 or more, a claim about its own demonstration, not an independent count. Some newer electric car battery packs run at up to 900 volts, the reason the laboratory gives for keeping a person away from a pack once it has been picked up.
A figure Oak Ridge National Laboratory reports for its own demonstration line, not an independently audited or industry wide number. The laboratory states 100 as a floor, "100 or more," not a precise count.
Show the numbers
| By hand | 12 |
| Automated line | 100 |
The limiting factor is the time it takes to perform the electrical discharge and perform disassembly manually.
Jonathan Harter, project team member, Oak Ridge National Laboratory. Source 1.