SLAMcore Blog

Robots Get Lost

Posted by Owen Nicholson on Sep 13, 2019 10:30:00 AM
Owen Nicholson

A new age of robotics is dawning, and I believe that robots will play a pivotal role in overcoming the real existential threats to our world – global warming, scarce resources and ageing populations. Robots are our future, but we are further away from fully autonomous robots than we should be. The reality is that, outside of very controlled environments, most robots really struggle with the tasks humans take for granted - ‘simple’ things like being able to understand where they are, and to move safely around an environment without crashing or getting stuck.

In this series of blogs I will look at the technical and structural barriers holding us back from developing cost-effective autonomous robots, and illustrate how innovators like SLAMcore are helping to overcome them.

Slamcore-robot-mazze-1-04-1

The march of the robots

First, let’s look at the technical difficulties robots face in navigation.

There are already thousands of robots working around us, in factories, warehouses and hospitals. However, the vast majority are fixed and can only manage repetitive pre-programmed tasks. Mobile robots and drones are being introduced for inspection in difficult or dangerous environments – but these are usually controlled by a human. Truly autonomous robots that can move around unaided are very rare and very expensive. They require a set of technologies known as SLAM which stands for Simultaneous Localization and Mapping. This is the ability for a robot to figure out where it is and map the environment around it by ‘seeing’ walls, objects and obstacles so that it can navigate around them.

Expensive mistakes

Living creatures instinctively understand how to cope with changes to the physical space around them; we can tell how far away we are from different objects; we know what those objects are and when to turn or stop to avoid them. For a robot to do this, it must combine, analyse and process data from numerous sensors, create a real-time 3D representation of the world around it, and then take decisions about whether it is safe and possible to move into a volume of space. Unfortunately, even with cutting edge hardware, expensive trial and error testing and bespoke algorithms, most robots still struggle with things like changes in lighting, fast moving objects, or environments that change or are unfamiliar. The most common reason for breakdown of autonomous robots is a failure of localization – they get lost and stop.

Currently, the robotics industry is struggling to overcome these issues using many different sensors, from contact sensors, to LIDAR, cameras to GPS. But every sensor adds cost, weight, power-demand and complexity. Plus, all these elements need to work together within the physical constraints of the overall robot design. More importantly, all the data from these sensors must be analysed, combined and processed to enable the robot to decide where to go. This needs to be done in real-time, and for a truly autonomous robot, on board! This in turn requires expensive, power-hungry, bespoke microprocessors and highly complex algorithms and software.

Suddenly, faced with these costs and the design complexity, the commercial case for a mobile, autonomous robot evaporates.

So, the march of the robots is on pause while they figure out where they are, where they are going, and which route they will take to get there.

Eyes&Ears

Use your eyes

But things are changing. Advances in AI and machine learning are helping us create software solutions that we thought were decades away. The falling prices of commodity electronics is making it easier to create standardised hardware modules. Counterintuitively, the secret to effective SLAM is fewer sensors not more. Taking inspiration from biology and the way we locate and navigate using just two eyes and the inner ear’s capability to detect motion, SLAMcore has created a module for simultaneous localization and mapping using just two cameras and a sensor that measures acceleration and rotation. These low-cost sensors are created by the million for smartphones and are readily available, so SLAMcore’s software solution can be easily embedded into almost any robot.

 

As I will discuss in the next blog, this ability to work across any number of mobile robot designs whether they walk, roll, fly or swim, will transform the whole robotics industry. Innovative start-ups with great ideas for robotic solutions will be able to focus on bringing them to market quickly and cost effectively, knowing that the thorny issue of localization and mapping has been solved for them.

Overcoming location and navigation issues is critical. SLAMcore has taken great strides in doing so, now the industry can forge ahead and deliver the robotic solutions we need.

Over the next 12 months we are going to be making some exciting product announcements. If you want to be the first to know, sign up to our early-access list.

 

Topics: VSLAM, RobotsForGood, SpatialAI, AI, Robotics