Hot pot money + an old mobile phone = autonomous navigation robot?

However, openbot, which Intel Labs recently opened up, costs only $50 and an old phone pulled out of a drawer. < / P > < p > the “brain” of openbot is an Android smartphone, while the “body” is an electric car built for $50. In addition, the researchers have developed a software stack that enables the smartphone to move the car. It seems easy to make, but openbot can actually support high-level robot tasks, such as human tracking and real-time autonomous navigation. < / P > < p > first of all, as long as you have a smart phone, you can meet three requirements: processor, sensor and communication interface. At the same time, the production cost of robot body is quite reasonable. If you disassemble them, you find that they are almost all conventional assembly parts: four TT motor wheels, one L298N motor drive plate, batteries and battery holders, screws and nuts, springs / rubber bands, wires, etc., and maybe you need two orange LED lights. < / P > < p > the total cost of the whole fuselage is only $50. 3D printing model files are also open source and can be found on GitHub. < p > < p > after testing 10 different types of smart phones, the results show that the openbot can always stably complete the human tracking task. < / P > < p > in addition to the cost, there are many advantages in choosing a smartphone as the “brain” of a robot. Hardware components on custom robots can quickly become obsolete. Consumer smartphones iterate every year, resolution and camera frame rate are improving, processors are faster, and sensors and communication interfaces are updated. < / P > < p > the advantages of smart machines are not only the rapid improvement of hardware capabilities, but also the booming software ecosystem. The research uses a mobile physical fuselage and software stack to enhance the robot’s perception and computing capabilities. < / P > < p > a small electric vehicle is designed based on cheap and available components. The hardware cost is only $50. This electric car can be used as the “fuselage” of wheeled robot. < / P > < p > the study developed a body for the robot based on available electronics and 3D printed parts for a total cost of $50, 40% of which was spent on batteries. If multiple robots are built, the cost will continue to fall. The details and prices of all parts are shown in Table 1 below: < / P > < p > Android applications run on smart phones, which can provide an operation interface, collect data sets, and run high-level perception and control tasks. The Arduino program is responsible for low-level drivers and metrics. < / P > < p > this study compares this wheeled robot with existing robot platforms, as shown in Table 2 below. Compared with other robots, the platform has a lot of processing power, communication interface and sensors provided by smart phones. < / P > < p > existing robots usually rely on customized software ecosystem, which requires special lab personnel to maintain code and implement new functions. And the research uses Android, one of the largest and growing software ecosystems. Without much effort from the robot community, all the low-level software for sensor integration and processing already exists and continues to improve. All sensors have achieved clock synchronization, thus eliminating many of the major challenges faced by existing robots. < / P > < p > this experiment studies the feasibility of running complex AI models on smart phones. The researchers used target detectors and applied visual servos to track people. Experiments show that all the latest midrange smartphones can continuously track people at 10 FPS or higher. < / P > < p > the researchers trained a driving strategy that works in real time on most smartphones. The strategy learned is to move along the corridor and turn at the intersection. The researchers compared it with existing driving strategies and achieved performance similar to that of the baseline, while the required parameters were reduced by about an order of magnitude. < / P > < p > in addition, the study has successfully transferred driving strategies to different smartphones and fuselages. The experimental results show that the driving strategy of this study can be generalized to the environment, new objects and even dynamic obstacles, even when there are only static obstacles in the training data. < / P > < p > Amazon sagemaker is a fully hosted service that helps developers and data scientists quickly build, train, and deploy machine learning models. Sagemaker completely eliminates the heavy work of every step in the machine learning process, making it easier to develop high-quality models. < / P > < p > now, enterprise developers can get a 1000 yuan service voucher for free, easily start Amazon sagemaker, and quickly experience five artificial intelligence application instances. Yueshang group has become the third social e-commerce service platform listed in China after being gathered in pinduoduo