As service robots are applied to more and more life scenarios, the complexity of the scenarios and terminal requirements are also increasing significantly, and the performance requirements for robots are getting higher and higher. In the important opportunity period of the current robot industry upgrade iteration, new A round of reshuffle is inevitable. In the collision of new demands and new products, who can become the new industry unicorn?
Under the escalation of demand, the problem is difficult to solve
Under the catalysis of the epidemic, service robots have ushered in unprecedented development opportunities, and the robot market in various segments such as cleaning, navigation, and food delivery has grown rapidly. However, at the same time, negative voices about robots have never stopped. Intelligence and security issues are the main pain points:
1. Lack of security
Taking the common supermarket scene as an example, it is densely populated, highly mobile, and contains many dangerous scenes such as narrow passages, glass guardrails, escalators, etc., which poses a huge challenge to the robot's obstacle avoidance and recognition capabilities, and the actual safety performance is not ideal. , such as the falling incident of the service robot escalator that occurred before. In the case of increasingly complex application scenarios, security capabilities are undoubtedly an important factor affecting commercialization.
The safety functions of most service robots currently have the following main disadvantages:
• There is no comprehensive security strategy, and the overall system linkage and scalability are poor;
• Only several fixed security scenarios are supported, each scenario is processed separately, and the security coverage is low;
• Serious dependence on sensors and high cost.
It should be mentioned that lidar is difficult to effectively detect highly transparent objects such as glass. Most of them require technicians to manually identify the position of the glass during the mapping process and delineate virtual walls in the map, so that the robot can complete the planning and Movement, resulting in reduced convenience of the robot.
2. Smart Geometry? Gimmick is bigger than reality
Since the birth of the robot, although it has gone through many iterations, the hat of artificial "mental retardation" has always followed the robot. In addition to being affected by excessive marketing, the key is still the slow progress of intelligent technology. It can be seen that most of the robots currently on the market have the same defects in terms of intelligence:
• Poor perception & cognition
Limited by the sensor characteristics of lidar, the lack of environmental semantic information makes the robot inherently deficient in recognition ability and scene understanding.
• Low level of intelligent decision-making
To "develop intelligence", a robot must not only be able to "see" and "understand", but also "understand execution". cognition) and knowledge-based decision-making.
• Unable to interact naturally
The interaction method is single, the response is relatively rigid, and the agility and functionality are insufficient.
It is foreseeable that in the future reshuffle, solving key issues such as intelligence and security will be the core of establishing future product competitiveness. From some existing products, it can be seen that some companies have achieved certain results in this direction, but overall it is not obvious. The reason is that the technical threshold and investment in intelligence and security are extremely high, making this Robotic companies facing financial pressure are somewhat stretched. Even traditional giants with deep pockets are faced with building a team and R&D system from 0 to 1, which is a huge investment.
So, how should companies deal with this problem?
Facing the dilemma of self-research, choosing to cooperate with AI suppliers may be an effective choice.
After years of development, the value of suppliers in the upstream of the industrial chain is becoming more and more prominent.
As a leading domestic supplier of key AI technology for robots, INDEMIND has rich accumulation in key technologies and product development such as robot navigation, obstacle avoidance, decision-making, AI interaction, etc., and has a full stack for robot companies to build products from 0 to 1 technical capabilities. Aiming at the intelligentization and safety of robots, INDEMIND has developed a systematic intelligent decision-making technology system, which realizes various functions such as intelligent obstacle avoidance, active safety, intelligent operation, and decision-making interaction. And according to the needs of robots in commercial scenarios, technology integration was carried out, and the RBN100 commercial robot AI solution was specially launched.
The solution is based on the INDEMIND OS Fusion system and adopts a multi-sensor fusion architecture to meet the development of core functions such as navigation and positioning of commercial robots, intelligent obstacle avoidance, path planning, and decision-making interaction.
To solve the problem of intelligentization and safety of robots, multiple technologies need to be connected and integrated, but the most important thing is the improvement of perception & cognition and intelligent decision-making.
Therefore, in terms of perception & cognition, INDEMIND adopts a multi-sensor fusion architecture with binocular vision as the core, supports the integration of mainstream sensors of different categories on the market such as Lidar, TOF, drop, and collision, and greatly improves information perception capabilities. Combined with VSLAM Algorithms that can realize real-time map construction of 3D scenes. At the same time, with the help of the leading AI recognition algorithm, it also supports the recognition of people, animals and various scene semantic recognition, and can effectively recognize and detect common glass doors (including glass turnstiles), glass guardrails, escalators and other scenes in commercial scenes.
In terms of intelligent decision-making, the self-developed INDEMIND intelligent decision-making engine technology can realize various functions for robots such as intelligent obstacle avoidance, active safety, intelligent operation, and decision-making interaction.
• Intelligent obstacle avoidance
Support low obstacle avoidance (ground socket, table and chair base, etc.); high reflective obstacle avoidance (glass, mirror, etc.); full height obstacle avoidance (desktop, warning line, etc.), to ensure the safety of the robot in complex environments run.
• Active safety
When facing a highly dynamic environment, the robot can make real-time potential risk judgments (pedestrians, pets, and fast-moving objects), and make avoidance strategies in advance according to risk classification, which is both reliable and flexible.
• Smart work
According to the dynamic operation requirements in the scene, the operation strategy can be adjusted independently to optimize the performance and efficiency of the robot operation.
• Decision interaction
Accept the natural semantic commands issued by users, independently design operation strategies, and complete personalized interactions.
Based on the powerful cognition & perception based on vision, combined with the systematic robot intelligent decision-making system, the intelligence and safety performance have been fully upgraded, and the user experience has also been improved across generations, so that it has the reliability of industrial-grade robots and the performance of small robots. Flexibility and higher cost performance at the same time provide an effective choice for solving the technical needs in the current industrial upgrading iteration.