Autopilot vs Indoor Positioning Explained | Marvelmind

0:01 Hello colleagues. Let's discuss today the difference between autopilot and indoor positioning system, because we receive quite many questions: can we program your indoor positioning system to control our robot? Well, the answer shall be obvious, but it seems that it is not. So let's do the basics. First of all, any autonomous system, including robots and us, includes three major parts: sensors, brain, which is processing and memory, and actuators. If you have these three elements, it's an autonomous robotics system and let's say intelligent system in general. So if you translate this to indoor positioning system, then our indoor positioning system actually belongs to sensors. So it sends the location and gives the input information.

0:58 To the autopilot, which is the brain. The brain itself is not capable to drive your robot or move your copter. So the motors and rotors actually do, but they are stupid. So they need a control, and this control comes from autopilot. So this is why the answer is absolutely basic and obvious and shall be very, very clear. Autopilot is a brain which makes the decision where to drive, how to drive, how to overcome the obstacles based on the sensor information which is coming from different sensors. Marvelmind indoor positioning system is one of those. You can have many—you know, magnetometer. You can have lidar for obstacle detection and avoidance. You can have odometry. You have many, many different types of sensors—barometer, if we're talking about drones.

1:56 So those are sensors. Yes, yes, in some cases, like autopilot like Pixhawk board, it already has some sensors, like IMU. So gyroscope is there, accelerometer is there, as well as our beacons. It's there. But you must very clearly distinguish between sensors, brains or autopilots, and actuators, which are motors and rotors. So let's do the same—like a summary. So autopilot controls—like physically controls, makes the decision—your actuators, your motors, your rotors, your whatever—what is doing things, what is moving your thing. It doesn't have any idea about the location. So this is why, of course, autopilot is important. Of course, we are important. But only together we do the real system. Stand-alone? No. You need input. In basic Pixhawk setup, it would be GPS.

2:54 You fly indoors. There's no GPS. So you need some sensory input about, "Oh, where am I?" Okay, we give that out of the box. At the same time, you know, returning to the original question, we do provide a real-time indoor positioning system and indoor positioning localization. Sometimes we kind of call it indoor-plus. So you can set waypoints using us. You can get the data from your device and send data to your device using our system. So you cannot do this through GPS because you don't have access to GPS, but you can use our system for that. So we are kind of indoor positioning system-plus, or GPS-plus-plus, from this perspective. But we do not control your motors directly. We send the data, and autopilot makes the decision where to go. Okay, I'm off the line. I need to go there. But it's not we that make that decision. Autopilot does.

3:53 So there are many different types of autopilot. Okay, the most common—and I'm using this as a kind of an example for all kinds of projects—it's Pixhawk-based. You know, Pixhawk is for me like Xerox, all kind of copying machines. So Xerox in this case. Pixhawk is just you know, a manifestation of any type of autopilot. So when I'm saying Pixhawk, I don't mean exactly Pixhawk. I mean any kind of autopilot which consists of the hardware and the stack. And for example, in case of Pixhawk, there are even two stacks: ArduPilot and PX4. And they are kind of independent, with some pluses and minuses, but there are two independent stacks on the same hardware. And you can write your own stack, autopilot stack, on the hardware. We also have our own autopilot. Okay, Marvelmind autopilot. We don't have it as a product for you, but of course we have autopilot in our own robots and for drones. For example, we can fly DJI drones.

4:53 So we also have autopilot. And in our perspective, autopilot is the application which runs on their phone—in this case, Android phone—which is connected together with DJI remote control. So that is autopilot. It's a primitive one, from waypoint to waypoint, but that's exactly what customers want. So they want the drone to be flying something autonomously, doing you know some task like taking pictures. But that's what autopilot does. Based on what? Based on localization data: where am I? Means, where am I, the robot or the drone? And the waypoints. The waypoints may come from either the artificial intelligence of the robot, if it's completely autonomous, or typically it comes from you as a user. Okay, I want to fly this warehouse on this pattern, or I want you to drive this pattern.

5:52 So you click waypoints, then autopilot knows, "Okay, I'm in this location because the indoor positioning system from our Marvelmind gives this location. I need to drive there. Okay, I need to turn because I have two, for example, mobile beacons, and I know my pose, so I can drive there directly. Okay, I go, go. Oh, okay, I'm shifting a bit. Okay, I need to turn." That's all what autopilot does. So this is autopilot decision-making, but based on our indoor positioning system. Otherwise, the autopilot has no clue where it is. Of course, it can use other systems. I'm not saying no. Sensor fusion is the best. It can be optical. It can be odometry. Of course, IMU. IMU in autopilot or let's say in autopilot hardware like Pixhawk, but also in our own beacons. They all have IMU. So they can get the data from both directions and then you know fuse and utilize the maximum, you know, drift, noise, all kind of things.

6:50 And it's always nice to have a comparison between two different systems. So, what are the key elements of autopilot and indoor positioning system? So a flight controller—or let's say drive controller—is hardware, you know, it's a brain, physical brain. So you must run the software or stack somewhere. And in this case, it's typically a small board. Depending on the applications, this board may have additionally, or let's say, additional sensors already. So it's autopilot-plus, from this perspective. But not necessarily. So at least it must have a processor and a memory to make the decision, and all the sensors may be external. But it's not necessarily like that. So often some sensors are already installed, like IMU, because they are small, cheap, et cetera. Then you have all these sensors I already mentioned and electronic speed controller. Those can be external, or they can be on the same board.

7:49 So they're kind of in the middle because they may belong to actuators already, kind of interpreting logical signals to the physical signals or data to control the motor or to control the direction. So, but it may also be a part of autopilot board or part of autopilot hardware. If we're talking about indoor positioning system, it's typically comprised of three major parts: stationary beacons or anchors, depending on what technology you're using. Mobile beacons, which you put on your mobile device, and a kind of central controller, which typically does synchronization inside the system, kind of controlling the system and sending the data to the external world. For example, in our case, you can get the location data directly from the mobile beacon, and it's typically recommended for autonomous vehicles because you don't need to run to the modem, then from the modem to whatever computer, then from the computer to this. Sometimes it's possible, but then it adds latency.

8:48 But sometimes the brain of your robot may be outside, so your robot may be smart, but the smartness is in the external computer. Okay, then in this case, it's easier to get the data from the modem. So we give you this flexibility. But in general, all indoor positioning systems consist of these three major elements: stationary beacons, which are kind of your satellites—GPS satellites—what you measure location against. Okay, against what? You are measuring the location—against. And mobile beacons, which is what exactly is measuring the location of your mobile device. Because we don't measure the location of your robot. We don't measure the location or position of your drone. We measure the location of your mobile beacon on the robot. And then we of course assume that it's not moving, so it's not shaking there. So then you know exactly the point which we are measuring.

9:48 Against the robot. And when their accuracy is very high, so it's also important that your Marvelmind is, you know, shifted from the center of your robot. So you must take this into account—that measuring this location, not, you know, central location or mechanical center or whatever physical center of your robot—no, we are measuring this. And even on our own beacons, for example, it's also important to understand that this is, for example, if you're talking about Inverse Architecture and receiving microphone, so this is the microphone, not this. This is a transducer. So this is kind of transmitting center, and this is receiving center. So for less accurate systems, like giving you 10-50 cm accuracy, it's not important. But for our system, which gives you centimeter level, of course it's already important where exactly the center is. So this is the center if it's receiving, and this is the center if it's transmitting. So that's a very, very quick, short, and hopefully easy explanation of the difference between autopilot and indoor positioning system.

10:47 Autopilot controls your robot or drone. They control your motors directly based on the input data which is coming from the indoor positioning system—basically localization—to drive from point A to point B or to fly from point A to point B. Thank you very much. Bye-bye.