Robots have a wide range, including environmental perception, decision-making and planning, motion control, human-computer interaction, etc. For undergraduate CS, it is recommended to focus on perception, such as SLAM. Common SLAM methods include laser, vision, and multi-information fusion. Whether it is research or employment, this direction is more popular. Perception is equivalent to human eyes. There is also decision-making and planning, which is the next link after the perception module. It is also very promising to do object prediction, path planning, and trajectory planning, with the same prospects as above. Decision-making and planning is equivalent to the human brain. Finally, there is motion control. There are many advanced control algorithms in academia (watered-down), but the actual application is still PID, or even PD, these classic algorithms. If you want to develop in the field of industrial robots in the future, you can still consider it.

Focus on learning embodied intelligence, that is, the combination of LLM and humanoid robot. In the future, within the scope of power supply, all human operation work will be replaced by robots.

For practitioners in the robot industry, this answer is biased towards robots with hardware entities.

1. Application of large language model (LLM), the most popular and the most popular collaborative robots, educational robots, and service robots are hotly researched. This is due to the launch and promotion of OpenAi represented by ChatGPT. The combination of large language models and subdivided industries has just started. The market is hot but talent is scarce. This is a direction.

2. Embodied intelligent humanoid robots are highly praised. It can be simply understood as robots of various forms, allowing them to perform various tasks in real physical environments to complete the evolution of artificial intelligence. However, embodied intelligence is very dependent on the technical development of basic disciplines.

3. L-SLAM navigation, V-SLAM navigation intelligent driving, AMR mobile robots, service robots, humanoid robots and other core technologies related to navigation and path planning that the robot industry continues to pay attention to. Generally speaking, these algorithms and technologies are combined with ROS. The hottest event for this technology is Wuhan Carrot Travel, where the car sensor was blocked by a plastic bag and the car could not run. SLAM is the core technology hotspot at present and for a long time in the future.

4. Visual recognition algorithm and radar recognition algorithm are the previous core of the robot navigation algorithm. Different algorithms are also closely related to different sensor devices, which belong to the basic scientific nature of the robot industry. This type of talent has a wide range of employment opportunities, not just in robotics. 5. Sensors such as lidar, binocular cameras, depth cameras, ultrasound, and TOF sensors are important technical cores of robot embodied intelligence. This is a more basic part that is more likely to subvert the robot technology ecosystem. At present, almost all good sensors in the industry are foreign brands. In recent years, the localization of sensors has been in the spotlight, and many companies have also taken advantage of the trend to rise rapidly. 6. Batteries are not just in the robotics industry. For any equipment that uses energy storage, this is an unavoidable technical bottleneck. AlphaGo and Ke Jie need to consume one ton of standard coal to play a game of Go. It can be seen that energy and energy storage seriously restrict the further large-scale promotion of robots.

5. Application of computer science in the field of robotics Programming: Master programming languages ​​such as Python and C++ to develop software and algorithms for robots. Computer vision: Enable robots to understand and process visual information for navigation and object recognition. Machine learning: Enable robots to learn from data and improve their decision-making and prediction capabilities.

6. If you majored in computer science and technology in your undergraduate studies and are interested in robotics, you can check out the following references: 1. Strengthen programming skills Programming is the core skill of robotics. You need to be proficient in the following programming languages: Python: concise and easy to learn, with strong library support. C++: suitable for control systems and embedded systems with high performance requirements. ROS (Robot Operating System): an open source framework developed specifically for robots. 2. Learn control theory Control theory is the basis of robot motion and behavior control. You need to master: Classical control theory: such as PID control and state space analysis. Modern control theory: such as linear system theory, nonlinear control and optimal control. Control of sensors and actuators: understand the working principles and control methods of various sensors and actuators. 3. Master machine learning and artificial intelligence Machine learning and artificial intelligence technologies are widely used in robots. You need to focus on: Supervised learning and unsupervised learning: understand linear regression, decision trees, support vector machines and clustering algorithms. Deep learning: master neural networks, convolutional neural networks (CNNs) and recurrent neural networks (RNNs). Reinforcement learning: familiar with Q-learning, deep Q networks (DQNs) and policy gradient methods. 4. Study robot kinematics and dynamics Robot kinematics and dynamics are the key to achieving precise motion control. You need to master: Forward and inverse kinematics: Learn how to transform from joint space to workspace. Dynamics: Understand the force analysis of robots and the derivation of dynamic equations. 5. Understand sensors and embedded systems Sensors and embedded systems play an important role in robots. You need to focus on: Sensor technology: Master the working principles and data processing of various sensors. Embedded systems: Learn programming and debugging of embedded systems. Participate in robot projects and competitions Theoretical knowledge is important, but practical experience is equally indispensable. Participating in robot projects and competitions can help you apply theoretical knowledge to practice. Here are some activities worth participating in: Robot competitions: such as RoboCup, FIRST Robotics Competition, DARPA Robotics Challenge, etc. Open source projects: Participate in open source robot projects, such as various projects in the ROS community.

7. Elon Musk said at the Tesla shareholders' meeting this year that the ratio of humanoid robots to humans may be about 2:1, that is, the demand for humanoid robots in the future may be 10 billion to 20 billion units, and there is still room for a significant price reduction from the target price of 20,000 US dollars. In the future, mass production will be ushered in in stages. However, due to insufficient application scenarios, coupled with high technology and high cost thresholds, humanoid robots are still in the early stages. At present, humanoid robots have relatively simple functions and have only achieved preliminary intelligence. In some entry-level application scenarios, they cannot compete with more mature and low-cost robots. At present, most humanoid robots focus on the development of two types of control modes, "position control" and "force control", and have not yet been deeply integrated with technologies such as deep learning and reinforcement learning. In commercial applications, stability issues still need to be overcome. In addition, one of the biggest bottlenecks is the high manufacturing cost, which depends on the foundation of large-scale production and multi-faceted technological breakthroughs, which further makes it difficult to promote humanoid robots. Take UBTECH, which is striving to become the "first stock of humanoid robots", for example. It claims to be the world's first company to reduce the cost of bipedal human-sized humanoid robots to less than $100,000, and has achieved the commercialization of the first large-scale humanoid robot in China. However, according to UBTECH's prospectus, UBTECH only sold one human-sized humanoid robot Walker-2 for education in 2021, and sold two Walker-X for general commercial purposes in the first nine months of 2022. From 2020 to 2022, UBTECH's Walker series revenues were 2.3 million, 12.8 million, and 51.85 million, respectively, accounting for 0.3%, 1.6%, and 5.1% of revenue, respectively. In contrast, consumer-grade robot hardware and solutions are the new growth point. From 2020 to 2022, the revenue share increased from 8.0% to 33.6%. How to promote the early maturity of humanoid robots and their large-scale application? The industry believes that the development history of smartphones can be used as a reference, that is, the early stage was mainly driven by hardware technology progress, and as the performance of the device improved and the functions increased, it brought more possibilities for content and application innovation, thus entering the stage of ecological development. Combined with this trajectory, core hardware should be the first area to break through. On the one hand, high-performance components can achieve better control and interaction capabilities; on the other hand, hardware solutions with redundant space can achieve high versatility and flexibility, which is conducive to further expanding the development and application ecology.

8. At present, the core components of humanoid robots mainly include motors, reducers, sensors, etc. Multiple motors are required at the joints of the robot. The motor drive system is a physical system that converts electrical energy into kinetic energy. It is mainly composed of loads, control devices and motors. The motor drive chip is the brain of the motor drive system. In humanoid robots, motor demand is widely distributed. Taking Tesla Optimus as an example, its main part uses 28 motor actuators to complete actions such as raising hands and bending knees. Sensors are mainly responsible for monitoring and interacting with internal and external environmental information. According to the source of information, they can be divided into internal sensors and external sensors. Internal sensors are mainly responsible for collecting their own motion and position information (such as geometric quantities such as linear displacement and angular displacement of joints, speed, angular velocity, acceleration, etc.), so as to achieve more accurate and reliable intelligent control. Compared with other mechanical equipment, robots have stronger interactivity with the external environment, so they need to use external sensors to monitor the surrounding environment parameters in real time to assist in target recognition, decision-making and judgment. In addition, there are 3D vision, wiring harnesses and connectors, etc., which form the eyes, nerves and blood vessels of humanoid robots. It is expected that in the early stage of the development of humanoid robots, it is necessary to vigorously develop general hardware such as smart chips, servo systems, and reducers, and as the demand increases, high-performance and low-cost standard solutions can be further formed. In addition, algorithms, as the core of humanoid robots, need to match the development of hardware. The current core problem lies in the control of algorithmic motion capabilities, including planning and control of body balance, walking gait, hand grasping, etc. Because humanoid robots need to complete various human-like actions, which are continuous, complex, require frequent physical interactions, and have many causal operations, the algorithm difficulty is even greater than that of autonomous driving. This requires a mature perception system foundation, powerful algorithms to decompose tasks and plan actions, continuous simulation training of large models, and super computing power support to form a continuous iteration of algorithm and hardware matching.

9. The computer machine learning engineers, data scientists, artificial intelligence ethicists, natural language processing engineers, robot designers, AI engineers, software development engineers, data privacy protectors, image recognition engineers, information security experts, artificial intelligence educators, data analysts, etc. created by artificial intelligence manufacturing will not only lead to a group of people losing their jobs, but also drive some people to employment. Tesla robot, now the humanoid appearance is completely enough. There is no need to develop in the direction of being more like a super artificial intelligence humanoid, otherwise it will begin to move away from the first principles of physics that have been adhered to. The application scenario can develop into the most dangerous area operation. For example, saving people in a fire scene, deep water operations... It is too dangerous for human life but has to go to the human field to fulfill its mission of existence, how about it? I recently learned a new course called communicating with bots. In addition to studying the latest AI robots, such as ChatGPT, Co-Pilot, and Gemini, I also need to master some skills and try to develop relatively simple robots related to electronic communications that I need. It is quite meaningful. The constantly updated and evolving courses feel in line with the times. It is better to go to a third-tier university or a junior college to study for a master's degree. However, these latest courses are more useful than what I learned in my undergraduate studies.

For more informtion, you can visit Robotic Search Engine to find latest research papers, AI Agent Marketplace and AI Agent Search Engine to find AI Agents related topics reviews and latest product information.