The development of AMR also depends on the progress of AI technology. AMR integrates perceptual recognition technology, big data mining technology, digital twin technology, and the technology of interconnection of everything.
Categorizing these technologies, we can find that the application of ARTIFICIAL intelligence technology in AMR mainly exists in three levels.
The first level is computational intelligence.
By drawing lessons from human experience and natural laws, the computer implements such intelligent optimization algorithms as tabu search, simulated annealing, particle swarm optimization, ant colony optimization, and genetic optimization, so as to deal with complex computing work such as task scheduling, path planning, pattern recognition, and data mining more intelligently.
The second level -- perceptual intelligence.
Obtain external environmental information through various sensors such as cameras, radar, and microphones, and then process the perceived information into results (such as identification and positioning) through image recognition, signal processing, and data fusion technologies, which serve as the basis for subsequent operations.
The third level is synthetic intelligence.
On the basis of perceptual intelligence, closely cooperate with business requirements, real-time adjustment of various AMR (including latent AMR, material box access AMR, composite manipulator AMR, forklift AMR, etc.) specific actions, to achieve this kind of high performance, high efficiency, high concurrency, low risk, low-cost logistics execution.
We have also done a lot of concrete work at different levels.
In terms of computational intelligence, we use an AI algorithm to make AMR achieve higher performance and faster path planning.
For example, we improved the A * algorithm, the traditional A * algorithm is not considering the interaction between many sets of AMR, we can make it improve dynamic body, according to the area of the vehicle running real-time vehicle routing planning, to implement the dynamic transformation of path direction, real-time dynamic obstacle avoidance, decrease the number of turns, minimize conflicts path number and the whole parallel execution time.
The application of computational intelligence in the AMR scheduling systems will help robots in three aspects.
The first is to achieve broad compatibility.
In the same map, compatible with different brands, types of robots;
In addition, it also supports the coordinated unified scheduling of multiple models such as TWO-DIMENSIONAL code models, SLAM models, pallet models, and material box models.
Then there is ultra-high performance.
Intelligent cooperation in path planning, load balancing, optimal task matching, job optimization, and other aspects;
Support efficient coordination of more than 1000 robots.
Finally, rapid deployment.
We adopt such modular and standardized system products to provide one-stop system solutions for planning, implementation, operation and maintenance.
In addition, when deploying, the project can be quickly landed just by updating the map.
The use of perceptual intelligence in AMR scheduling systems is also obvious.
The most typical perceptual intelligence is SLAM navigation, including laser SLAM and visual SLAM.
The ultimate goal of applying comprehensive intelligence to the AMR scheduling system is to achieve AMR solutions with high flexibility, high reliability, and high-cost performance.
High flexibility, mainly can adapt to a variety of sizes and layouts, rapid deployment and adjustment; Compatible with different business requirements. Meet a variety of modes of mixed running; Areas, equipment, and processes can be quickly and flexibly expanded.
In terms of high reliability, it mainly realizes environment perception, flexible adaptation, and reduces the failure rate. The system collects the health degree of the equipment remotely and pushes the maintenance plan. Hot-swap is mutually available, and a single failure does not affect the overall system operation.
High-cost performance, mainly intelligent scheduling, human-machine collaboration to improve the overall performance of the system; Rapid deployment and high reliability reduce implementation, operation, and maintenance costs; Flexible function adjustment, reduce development switching cost.
