With the needs of social development and the expansion of robot application fields, people's requirements for intelligent robots are getting higher and higher. The environment in which intelligent robots are located is often unknown and difficult to predict. In the process of studying such robots,

It mainly involves the following key technologies:

Multi-sensor information fusion

Multi-sensor information fusion technology has been a very popular research topic in recent years. It is combined with control theory, signal processing, artificial intelligence, probability and statistics.

It provides a technical solution for robots to perform tasks in various complex, dynamic, uncertain and unknown environments. There are many kinds of sensors used by robots.

According to different uses, they are divided into two categories: internal measurement sensors and external measurement sensors. Internal measurement sensors are used to detect the internal status of robot components, including: specific position and angle sensors; arbitrary position

, angle sensor; speed, angle sensor; acceleration sensor; tilt angle sensor; azimuth angle sensor, etc. External sensors include: vision (measurement, recognition sensor), touch (contact, pressure sensor)

, sliding sensor), force sensor (force, torque sensor), proximity sensor (proximity, distance sensor) and angle sensor (

Multi-sensor information fusion refers to the integration of perception data from multiple sensors to produce more reliable

, more accurate or more comprehensive information. The fused multi-sensor system can more completely and accurately reflect the characteristics of the detection object, eliminate information uncertainty, and improve information reliability. The fused multi-sensor information has the following characteristics

: Redundancy, complementarity, real-time and low cost. Multi-sensor information fusion methods mainly include Bayesian estimation, Dempster-Shafer theory, Kalman filtering, neural network, wavelet transform, etc.

。

Multi-sensor information fusion technology is a very active research field, and its main research directions are:

1 Multi-level sensor fusion Due to the weaknesses of uncertainty, observation error and incompleteness of a single sensor,

Therefore, single-layer data fusion limits the system's capabilities and robustness. For advanced systems that require high robustness and flexibility, multi-level sensor fusion methods can be used. Low-level fusion methods can fuse multi-sensor data;

Intermediate-level fusion methods can fuse data and features to obtain fused features or decisions; high-level fusion methods can fuse features and decisions to reach the final decision.

2 Microsensors and Smart Sensors The performance, price and reliability of sensors are important indicators of the quality of sensors.

However, many sensors with excellent performance are limited in application market due to their large size. The rapid development of microelectronics technology has made it possible to manufacture small and micro sensors. Smart sensors integrate main processing, hardware and software.

Scientific company developed 1000 series digital quartz intelligent sensor, Japan Hitachi Research Institute developed olfactory sensor that can identify 4 kinds of gases, American Honeywell

The DSTJ23000 intelligent differential pressure sensor developed by us has a certain degree of intelligence.

3 Adaptive Multi-Sensor Fusion In the real world, it is difficult to obtain accurate information about the environment, and it is also impossible to ensure that the sensors can always work properly. Therefore, for various uncertain situations,

Robust fusion algorithms are very necessary. Some adaptive multi-sensor fusion algorithms have been developed to deal with the uncertainty caused by the imperfection of sensors. For example, Hong proposed an extended joint method through innovative technology.

It can estimate the optimal Kalman gain of a single measurement sequence filter. Pacini and Kosko also studied an adaptive target tracking fuzzy system that can be used in mild environmental noise.

It incorporates the Kalman filter algorithm in the processing.