Welcome to Galaxis Consulting hotline:0086 400-101-0947

In-depth conversation with Dr. Bai Hongxing from Galaxis: AI-enabled AMR for high flexibility, high reliability and high cost performance?

The changing market demands rapid deployment and migration of robots, ultra-large-scale scheduling, and efficient human-machine collaboration. When AGV cannot meet the demands of a whole new logistics business, AMR technology comes into being.

When the logistics industry is heading for a rapid growth period, the market is putting forward new requirements of being more, faster, better and more economical to the logistics link. In general, high flexibility, high reliability and high cost performance are the challenges that all logistics link robotics companies have to deal with.

The AMR model is helping the rough and expansive primary logistics chain to shift to a refined, efficient and flexible logistics operation model.


Bai Hongxing, Vice President of Galaxis Technology Group, shared around three major issues.

 "What is AMR? What AI technologies does AMR use?”

"What are the current pain points of the logistics industry? Why do we need to achieve high flexibility, reliability and cost effectiveness?”

“How can AMR's AI technology help customers create a highly flexible, reliable and cost-effective logistics model?”


Bai Hongxing mentioned that the changing market demands rapid deployment and migration of robots, ultra-large-scale scheduling, and efficient human-machine collaboration. When AGV cannot meet the demands of a whole new logistics business, AMR technology comes into being.

He said that the current AMR technology is combining with AI technology to achieve new progress, and smart optimization algorithm scheduling technology, sensory recognition technology, big data mining technology, digital twin technology, and the technology of Internet of Everything is empowering AMR robots.


01 Why is AGV replaced by AMR?

AMR was developed from AGV technology.

With traditional AGV deployment, machines are defined before providing services, and the scale of multi-machine collaboration is often easily limited by the scheduling system.

In the AMR scenario, the QR code on the ground is gone, the models are enriched, and collaboration between machines is possible. Each machine can achieve multiple functions such as pallet access, handling, goods-to-man, goods-to-manipulator, human-machine collaboration, etc. AMR also has the flexibility to change different functions such as goods-to-man and order-to-man depending on business needs.


Why did the AGV become an AMR?

Actually, what behind it is various changes of by the market.

First, the market is changing, goods are becoming increasing diversified, customization and order fragmentation are more common, and the formats of the logistics business are becoming richer and more uncertain. The result of these uncertainties is that the functions, order structures and business processes that were pre-defined when the robot was first put into production are likely to change as the customer's business evolves at a later stage This requires logistics equipment systems that are highly and increasingly flexible, and able to quickly adapt to a variety of different needs.


Second, distribution efficiency is increasingly demanding.

In the past, the wholesale model and distribution model are based on week, but when in the e-commerce model, distribution begins to be in a daily basis, and community e-commerce front warehouses will even .be calculated in hours and minutes This requires logistics information systems that are increasingly smart and equipment systems that operate more efficiently and without interruption.


Finally, the market is becoming increasingly competitive.

Although labor is getting more and more expensive, every logistics operator is seeking ways to lower the operating costs in a market of fierce competitive. This requires logistics equipment that are highly and increasingly flexible, and can adapt to various needs. In addition, it needs to be in large-scale, and the requirements for the degree of automation are getting higher and higher. The initial investment and later operation should be extremely cost-effective.


In general, changes in the market have put forward higher requirements for the AGV system——

Robots are required to achieve rapid deployment and migration, ultra-large-scale scheduling, and efficient human-machine collaboration.

Thus, AMR robots came into being, echoing the major three demands of the market.


What distinguishes AMR and AGV most?

We can compare the two in three aspects: real-time deployment, scheduling scale, and human-machine collaboration.

In terms of real-time deployment, AGV has high requirements for site implementation and deployment, and it is not easy to adjust at any time.

AMR does not require coding, is highly adaptable to sites, and can be deployed within one day. AMR can flexibly change and increase or decrease according to site changes.


In terms of scheduling scale, AGV mainly realizes real-time centralized scheduling through the control system. As the number of AGV increases, the computational complexity of the entire scheduling system will exponentially multiply, and since real-time scheduling is required on site, the requirements for the communication network are very high. Once it fails to communicate with the central server, the AGV can only stay in place, send alarms and rely on humans to solve the problem.

In contrast, AMR can achieve distributed scheduling, with no limit of scale in theory.


In terms of human-machine collaboration, traditional AGV should minimize the intersection of people and AGV, which requires AGV to design safety fences, electronic fences, etc. at the interaction point, but AMR enables human-machine interaction anytime, anywhere.


02 How does AI technology empower AMR robots?

The development of AMR also relies on the progress of AI technology. AMR integrates perception and recognition technology, big data mining technology, digital twin technology, and the technology of Internet of Everything. Base on classification of these technologies, we will find that the application of AI technology in AMR mainly exists in three levels.


The first level - computational intelligence.

Computers learn from human experience, natural laws, etc. and implement smart optimization algorithms such as tabu search, simulated annealing, particle swarm, ant colony, genetics, etc. to handle complex computing tasks such as task scheduling, path planning, pattern recognition, and data mining in a smarter way.


The second level - perceptual intelligence.

The external environmental information is obtained through various sensors such as cameras, radars, microphones, etc., and then the perceived information is processed to form results (such as recognition, positioning, etc.) through image recognition, signal processing, data fusion and other technologies, which forms the basis for subsequent operations.


The third level - comprehensive intelligence.

On the basis of perception intelligence and in close coordination with business needs, the specific actions of various AMRs (including latent AMR, tote access AMR, composite manipulator AMR, forklift AMR, etc.) are adjusted in real time to achieve this kind of logistics execution featuring high-performance, high-efficiency, high concurrency, low risk, and low cost.


We have also done a lot of detailed work at different levels.

On computational intelligence, we enable AMR to achieve higher performance and faster path planning through AI intelligent algorithms.

For example, we have improved the algorithm. the traditional algorithm does not give enough consideration to the mutual influence between multiple AMRs. We improve it so that it can dynamically plan the vehicle path according to the real-time body of the vehicles running in the area, realize dynamic transformation of the path direction and real-time dynamic obstacle avoidance, reduce the number of turns, and minimize the number of path conflicts and the overall parallel execution time.


AMR scheduling systems adopting computational intelligence will help robots in three major ways.

The first is to achieve broad compatibility.

Robots of different brands and types are compatible in the same map.

It also supports the collaborative and unified scheduling of multiple models such as QR code models, SLAM models, pallet models, and tote models.


The second is ultra-high performance.

Smart collaboration in path planning, load balancing, optimal task matching, job optimization, etc.

Support efficient collaboration of more than 1000 robots of various types.


Finally, rapid deployment.

We use this modular, standardized system product to provide one-stop system solutions covering planning, implementation, operation and maintenance

In addition, when deploying, it is possible to get the project put into practice quickly by simply updating the map.


Perceptual intelligence also brings benefit to AMR scheduling system.

Among them, the most typical one is SLAM navigation, including laser SLAM and visual SLAM.

The ultimate goal of comprehensive intelligence applied to AMR scheduling system is to achieve a highly flexible, reliable and cost-effective AMR solution.


In terms of high flexibility, it adapts to various venues and layouts and realizes quick deployment and adjustment; it is compatible with different business needs. It meets the hybrid model needs; quick and flexible expansion can be realized for areas, equipment, processes, etc.


In terms of high reliability, it mainly realizes environmental perception, flexible adaptation and reduction of failure rate; The system remotely collects device health and pushes maintenance plans; Hot-swap and mutual backup; single-machine failure does not affect the overall system operation.


In terms of high cost performance, it is mainly about smart scheduling and human-machine collaboration improve the overall performance of the system; Rapid deployment and high reliability reduce implementation, operation and maintenance costs; flexible adjustment of functions reduce switching costs.


Finally, we would like to introduce the 6 major advantages of Galaxis’ AMR solution: 

First of all, we have mastered the core technology,  and all technologies are independently developed, including the bottom QR code technology, laser SLAM technology, visual SLAM technology, embedded control, autonomous positioning and navigation, etc. The ability of self-research enables us to be closer to the application needs of customers and engage in customization work to some extent.

The high stability of the equipment means reliability. High modularity of the components means it is relatively independent of structure and can quickly adapt to different models.

It has excellent expansion performance, can be equipped with universal bus, peripherals and sensors can be adjusted as required. In addition, our visual sensor has a large amount of information, and we can develop visual applications on demand.

Of course, our products also have shorter deployment times and short construction periods.

Finally, Galaxis’ overall solution is also very cost-effective thanks to the application of AI.


03 Q&As

Q: What is the lowest price range that a future robot can reach?

A: With the development of highly flexible, highly reliable and cost-effective AMR technology, some mainstream models will become popular. When annual production and sales reach hundreds of thousands or even millions of units, the price will be further reduced.


Q: In your opinion, will the robotics industry be dominated by one company in the long run? or will segmentation prevail?

A: It will not be dominated by one player; everyone familiar with the logistics equipment industry knows that, forklift industry has been developed for so many years, but no monopoly. For AMR, there are those specializing in pallets, totes, access, handling, etc. In addition, there are also different geographical attributes and industry attributes. So, monopoly is unlikely to occur. However, among the branches of mass models and niche models, giants are more likely to occur in mass model sector.

In some segments, there are bound to be some companies that do particularly well, make good profits and stand out. Of course, the total sales of these niche models will not be excessively large, and giants are more likely to occur in mass model sector.


Q: What do you think is the biggest obstacle to the development of domestic AGV robots now?

A: AGV are very costly to deliver, and there are many non-standard customizations. For individual projects, it involves software development, field deployment and commissioning costs.

The arrival of AMR brings more intelligence and flexibility. We can use fewer models to meet the needs of application scenarios, which will reduce the cost of non-standard customization, and at the same time, after the application of AI technology, software development does not need to be customized individually, which can be more productized and generalized. In the future, AMR will definitely have the function of OOTB (out-of-the-box), just like a floor sweeper.


Q: I saw on the official website that Galaxis made 12 scenarios, how did you make your robots adapt to the new scenarios quickly?

A: To find their commonalities after abstraction.

For example, for the type of materials handled, are they in pallet box type and are zero goods? or what are the impulse characteristics of these materials. For example, there are A class for high frequency, B class for medium frequency, and C class for low frequency. The function of the robot is mainly access, handling and sorting.

After extraction, we can find the commonality between different scenarios, plus the flexibility achieved by AI technology, when planning and designing the program, we can realize perfect match by considering the differences between different industries and different scenarios.


Q: Under what circumstances do you think AMR access robots can replace shuttle and stacker solutions?

A: A smart shuttle is also a kind of AMR, which runs on rails. Is AMR an alternative to AMR with tracks? I don't think so,  and the two will co-exist.

With rail, it will be faster. When there is high traffic, high throughput and higher access density, the shuttle will be more suitable; but in areas of lower stores with lower traffic requirements, AMR is better for flexibility.

Let’s take an example, that it, the car is now popular now, but we still need subways and high-speed trains.


Q: What do you think is the core competence of AMR in the future? Not just compared to AGVs, but in the whole logistics field?

A: AMR represents a kind of smart logistics equipment, and its biggest competitiveness lies in its strong versatility.

In the past, we talked about automated logistics equipment, but now it is smart logistics equipment.

In the era of automated logistics equipment, special machines are used for special machines, such as conveyor only for conveying, sorter only for sorting, and stacker only for access.

That AMR's versatility allows it to cross the functional barriers of access, handling, and sorting, and can adapt to more scenarios with better products, which is more flexible, more reliable, and more cost-effective, and it is where its competitiveness lies.


Q: Can you tell me the degree of reusability of Galaxis solutions in vertical industries? For example, what percentage of the logistics solution applied to one e-commerce customer can be replicated to the next similar customer? How can Galaxis provide relatively standardized products or solutions for non-standard scenarios?

A: We pay special attention to planning and designing solutions, and basically we don't use one-for-all solutions, because each customer's inventory characteristics and order structure are different, so we will design solutions based on strict data analysis. This process is also constantly improving our scheduling system, making our software products more versatile.


Q: They say AMR belongs to the fourth generation of mobile robots, what new technologies do you think will be integrated into mobile robots in the future? Are there some hints of technological directions for the industry?

A: AI is being integrated in, and we can focus on how to design some mobile robots for mass models. In addition to the current star products, we should think from the First Principle, such as how to achieve the separation and quick change of the chassis and top, to form the scale of the annual production of millions of units.


Q: In the market, there are also some ontology machine manufacturers in the market that focus on providing technology solutions for individual link such as robot controller. Is the emergence of such companies likely to significantly reduce the cost of robots?

A: Single link technology solution providers lower the threshold of being robotics suppliers rather than the cost of robotics.

Take controllers for example, to adapt to more customers, they have to be made more versatile, which may instead increase costs. However, for the controller made by the proprietor manufacturer itself and used in a particular model, only the necessary functions and interfaces are required, so the cost will be lower.

To significantly reduce the cost of robots, the most critical thing still lies in a versatile model or versatile chassis, and the key to reducing costs still lies in scale.