Autonomous Robot Navigation: Demo Coverage
Boxie 2 is a fully autonomous mobile robot engineered for industrial applications including barcode/QR scanning, 5G coverage optimization, and warehouse delivery. Weighing under 5 kilos, it delivers 2cm positioning accuracy using Marvelmind's ultrasonic indoor positioning system across areas as large as 100×100 meters. The major upgrade from Boxie 1 is automatic charging station capability, enabling truly continuous autonomous operation. Equipped with 12 LIDARs, omnidirectional sonar, IMU, odometry, and an up-facing camera, the robot combines multiple sensor fusion for obstacle detection and precise navigation on unpolished industrial floors.
Transcript
Boxie 2 is a fully autonomous mobile robot engineered for industrial applications including barcode/QR scanning, 5G coverage optimization, and warehouse delivery. Weighing under 5 kilos, it delivers 2cm positioning accuracy using Marvelmind's ultrasonic indoor positioning system across areas as large as 100×100 meters. The major upgrade from Boxie 1 is automatic charging station capability, enabling truly continuous autonomous operation. Equipped with 12 LIDARs, omnidirectional sonar, IMU, odometry, and an up-facing camera, the robot combines multiple sensor fusion for obstacle detection and precise navigation on unpolished industrial floors.
0:01 Hello, this is Boxie 2. Boxie 2 is our new autonomous robot for industrial applications. What kind of applications? Well, first of all, all kinds of scanning. So you install barcode, QR code, or cameras on the robot using their existing holes, and there are many on the top, on the side. It means that it's very convenient. And then you send the robot using waypoints from our Dashboard, and the robot drives repeatedly and accurately with 2 cm accuracy over an area as large as you wish. So, for example, if you have a warehouse of 100 by 100 meters, okay, not a problem. You install many stationary beacons, and then you send the robot driving. Or it could be even a more sophisticated scenario. For example, you install a 5G terminal on the robot, and the 5G terminal is
0:59 used for accurate 5G indoor coverage optimization. So it means that you tweak your network, and you send the robot. The robot carries the 5G terminal and even can get the power from the robot. Let's discuss a bit about this later because the robot has a lot of power internally and is designed to power your own equipment. So the 5G terminal is driving, or the robot is driving, carrying the 5G terminal. The 5G terminal performs calls, downloads the data, uploads the data, and you use this data. Then you tweak your network, and then you drive the robot again, and then you see whether there's any change. So, of course, in order to compare the first drive to the second drive, you need repetition. It's very difficult to do manually, but with the robot, it's exactly what the robot is designed for. But let's jump to the differences between Boxie One and Boxie
1:57 Two. Boxie One and Boxie Two are very similar sizewise. So the robot is around 4.5 to 5 kilos, depending on the battery configuration. So let's say under 5 kilos. But the biggest difference between Boxie One and Boxie Two is an automatic charging station. So the charging station is using the same 5 amp fast charger, which is capable to charge the robot fully with the default 100 watt-hour or 8 amp-hour battery in under two hours. So the same charger is connected to the charger, and the charger is automatic. So the robot comes to the charger. The charger detects there is a robot, and the charger puts electricity on the pad. So when there's no robot, there's no electricity, so you can't short circuit—not
2:56 a problem. But only when the robot arrives, it contacts the pins, and those pins will start getting their current from the charger. So this is the biggest difference because before the robot was fully autonomous, but now it's even more autonomous because basically it can drive forever until it breaks. Because it can drive automatically, but then when its battery is low, it automatically goes to the charging station and charges. So this is an autonomous robot. Of course, any autonomous robot must have sensors, actuators, and processing units. So let's talk about the sensors. Of course, the main sensor is the indoor positioning system, which is using our ultrasound plus radio-based system. So it has two omni microphones, and these omni
3:54 microphones, by the way, only top 25 cm are here. If you want to install a basket, for example, like a one meter basket, okay, not a problem, because there is a one meter cable inside. So it means that you can remove this, unscrew these four screws, and put it on top. So that means that your basket will not produce a non-line of sight issue. And the robot measures the location of each of its omni microphones with 2 cm accuracy. By knowing the location of two, it knows not only its own precise location but also direction, because in stationary positioning, it's not possible to get the location without having two omni microphones. Of course, omni microphones and the indoor positioning system is one of the sensors, but one of many. Even so, most important probably, but one of many. What are the others? No, first of all, lidars. Lidars are important for
4:52 obstacle detection and avoidance. There are 12 lidars altogether. Lidars can sense up to 4 meters, but then it would be too sensitive to external light. For example, so we set it to 1 meter, which is more than enough because the robot drives around 30 to 50 cm per second at maximum. So it means that it has two seconds to stop, which is more than enough. And the robot, and the lidar sends, and they stop, whatever 30 to 50 cm before the obstacle. It sends all around and also below. So the sensor is even below, so it means that it can sense a negative step. It will not fall from your stairs. Other sensors, now sonar combined with two omni microphones—it's very, very powerful because very often you may face, or the robot may face
5:50 this situation when lidars are not detecting. Like light on glass, for example. And also, light, by the way, lidars can be blinded with very, very powerful infrared. It's okay—we have a secondary source of information. It's not as precise, of course, as lidars, but it's better than nothing. And sensor fusion is the heart of the system. Then, of course, odometry. There is a very precise odometer on the left wheel and on the right wheel, and it measures their drive. When, for example, there's an occlusion and lidars are not sensed because the distance is very large, so it can drive for a few meters using purely odometer and still won't be lost. And then, of course, it accumulates the error, and this error is then canceled by the ultrasound system once again. A sensor-based or sensor fusion-based system. Of course, IMU. IMU is
6:50 used for everything again—for sensor fusion because without IMU, direction is not possible, or let's say precise direction is not possible, and many things are not possible. I guess there are even more sensors we can discuss. Just to make it a bit short, another very important element is the camera. There's an upward-facing camera, and those lights are not for nothing. So you can switch them on, and they would light up the ceiling. So it means that it's possible that the robot would be driving using an odometry and ultrasound-based system and optical positioning—for example, for QR codes, barcodes, and those special codes that you place on the ceiling. So it is also possible. Then, about the actuators, because the robot needs to perform something. The basic thing is, of course, driving, and that's basic, but the most important, probably. So yes, it drives
7:49 autonomously and very accurately—2 cm accuracy. And this 2 cm accuracy is provided by the sensors, but also by the wheels, which are designed for all this kind of driving in an industrial environment. For example, this surface here is not very smooth; it's pretty rough, so it's up to 5 mm or more. It can drive still successfully. So the floor must not necessarily be polished. What are the actuators now? Well, since you connect your own equipment, it's very important that you want to power your equipment. So it means that you can power this equipment from the external USB. It's very easy, like regular USB. You just connect it, and your external camera or external something will be powered. Then, of course, there's additional pins specifically designed to
8:47 power your equipment: 12 volts, two amps; 5 volts, two amps; and switched ground, also two amps. But you want to communicate with the robot as well. So this is why there is an open API. So it means that through the API you can get plenty of data from the robot and even send to the robot. What kind of data? No, first of all, of course, location. But then, of course, a lot of other pieces of information—like power supply, like how many meters it has driven, what is the current speed, what is the current consumption. Hundreds of different fields you can get. But even more importantly, very often you want to command your own
9:46 equipment—like switch on, switch off, focus, or do something. You can do this using our own system. So it's not very fast—a few kilobits or a few tens of kilobits per second—but you can use it. So it means that you can send from your system to your payload some commands using our own radio. You don't need additional LoRa or Wi-Fi or Bluetooth. You can send using our radio, and then through UART or SPI—let's discuss about this. You can command your own equipment, and vice versa. So it means that your own equipment can send the data from the equipment—like I don't know, position of their arm or that it took their measurement or it recognized the QR code—and it sends the QR code to the system. Okay, QR code is recognized. You use virtual UART over
10:46 USB, and then you send it through our radio. And you collect, or through the modem, using virtual UART to USB once again. So then, multiple interfaces. Let me repeat: virtual UART over USB, multiple USB ports, SPI, I2C. You get this connectivity using an odometry board, which is the low-level board, but also the higher-level board, which is basically a Linux computer—a Raspberry Pi. And Raspberry Pi is powerful because you can run even your own applications there, and you can debug those applications using the HDMI and USB. So you can connect the mouse and the keyboard and debug it. Additionally, there is Wi-Fi
11:44 and Bluetooth. So there are special holes on this. So it means that you can send the data using Wi-Fi. Okay, for Wi-Fi there's a cable, and for Bluetooth there's a small hole. So it means that you can collect pretty fast data out of the system. We do not recommend running your equipment or your program on their computer, but you can. Normally, you install your favorite board right on the top, and it's very convenient because you see there are many holes over there. But some of these holes are even designed specifically for Arduino, Raspberry, and Jetson. So it means that those holes are basically ready to use for your favorite boards. And you can run your application on your boards. You can supply your boards from the robot, and then you do all the things you want it to do.
12:43 By default, now it's about battery. By default, the robot contains 100 watt-hour batteries. So it's one, or 12 volts, and eight amp-hours. It's basically because we ship it using air, and there's a limitation of how much battery it can be in order to be allowed on the plane. But you can purchase additional batteries. So by default, they're eight amp-hours, but you can install up to 40 amp-hours—five times more. Eight hours is sufficient to drive several hours. It's very difficult to give a precise number because it depends on the load, on the speed, on the surface, on the mode of driving. But it's several hours with full capacity. Only with full capacity, it would be five times more, and our
13:41 estimated drive time would be 48 hours. Forty-eight hours. And let me repeat: so the default can be charged in two hours and drive around 8 hours or 10 hours. And then with the full capacity, it would be able to drive 48 hours. And, of course, typically you don't need so much for self-driving. So you power your external equipment from these huge batteries. Some people want to have it even more. Is it possible? Sure. There's a layer of batteries. So you install additional batteries, and then you just connect this battery directly to this 12 volts port. So it means that the robot will carry the battery, or even an uninterruptible power supply, and this power supply would be supplying your own equipment. So
14:40 what else about the robot? The robot is designed for industrial applications, but there are many other applications what people are trying to use the robot for. Now, for example, I already mentioned the basket. So originally it was for bringing samples of plastic. So there's a huge plant, and people produce plastic in one part of the plant, and the laboratory which needs to test and analyze the plastic every 30 minutes is on another side. So now people are driving every 30 minutes. Okay, what's the point? You install the basket again. You screw the basket. Put the bottle or with samples in it, and the robot drives back and forth, back and forth. So all kind of delivery up to
15:35 10 kilos. Ten kilos is the designed payload, but not only. Just recently there was a case, and people say, "Oh, we want to draw. Is it possible to use it to draw something on the floor?" Yes, of course. So the robot is very precise. Check their driving videos. And then, of course, you just install a pan or some painting device on this, and you control this painting device through either your software, or we can even write the application software for you. So it means that not only will the waypoints for the robot be sent to the robot, but also the commands for this painting device. Yes, we do produce application software or features based on the customer needs because everyone wants something special and something
16:34 unique. What we recommend: we recommend that you get the robot and start playing with it because we cannot imagine all kind of applications. Customers always have something even more special. So play with it. Tell us what you are missing, and then we will quickly introduce those additional features for you. And this way, we will make the robot even more suitable for your particular applications. And typically it's very quick—between a couple of days to a couple of weeks, depending on the complexity of the application. Boxie Two robot—a new mobile robot for industrial applications. We very much hope that you will enjoy it. Thank you very much.
Video Contents
- 0:00Use Cases: Scanning, QR/Barcode, and 5G Applications
- 0:535G Terminal Integration and Network Optimization Scenario
- 1:55Boxie 1 vs. Boxie 2: Key Differences and Automatic Charging
- 3:28Sensor Suite Overview: Microphones, LIDARs, and Odometry
- 4:48LIDAR Obstacle Detection and Range Configuration
- 7:07Up-Facing Camera and Optical Positioning with QR Codes
- 7:37Actuators and Autonomous Driving Performance
- 8:25Power Distribution: USB and Custom Power Pins
- 9:05Open API and Equipment Control Interfaces
Key Takeaways
- Boxie 2 achieves 2cm positioning accuracy across large warehouse areas using ultrasonic indoor positioning without line-of-sight requirements
- Automatic charging station enables truly autonomous operation—robot docks, charges fully in under 2 hours, and continues indefinitely
- Comprehensive sensor fusion combines 12 LIDARs, omnidirectional sonar, dual-wheel odometry, and 9-axis IMU for robust obstacle detection and navigation
- Open API and configurable power outputs (USB, 12V/2A, 5V/2A) enable flexible integration of cameras, scanners, and custom equipment
- Weighs under 5 kilos and navigates unpolished industrial floors with 5mm+ surface variations without requiring floor modifications
Relevant For: Robotics Engineers & AMR Developers
Warehouse managers, automation engineers, and facility operations teams seeking autonomous mobile robots for repetitive scanning, delivery, and positioning tasks. This content addresses the need for precise indoor navigation, extended autonomous operation through auto-charging, and flexible payload integration in industrial environments.
FAQ
Autonomous Robot Navigation System
Boxie 2 represents a significant advancement in autonomous mobile robotics for industrial facilities, combining precise indoor positioning with practical autonomous operation capabilities. The robot integrates Marvelmind's ultrasonic indoor positioning system—utilizing two omnidirectional microphones and radio-based beacons—to achieve 2cm accuracy across large warehouse areas without requiring polished floors or line-of-sight limitations. Its sensor suite includes 12 LIDARs for 1-meter obstacle detection, sonar for backup environmental sensing, dual-wheel odometry for occlusion handling, and a 9-axis IMU for directional accuracy. The upward-facing camera enables optical positioning via ceiling-mounted QR codes or barcodes. Boxie 2's transformative feature is its automatic charging station, allowing the robot to autonomously dock and charge its 100Wh battery in under two hours, enabling indefinite operational deployment. The robot supports payload integration through mounting holes and includes configurable power outputs: standard USB, 12V/2A, 5V/2A, and switchable ground pins. Its open API enables command-and-control via radio and wireless protocols, while supporting popular single-board computers like Arduino, Raspberry Pi, and Nvidia Jetson Nano. Use cases span warehouse scanning, 5G indoor coverage testing, inventory verification, and research applications.
Topics
Related Resources
📍 Need precise indoor positioning for your project?