These days, digitalisation in the outdoor industry cannot be separated from robotisation. But what are the challenges that arise when moving robots out of the factory and into the field? How can the possibilities be used to solve urgent problems in complex outdoor environments? What is the impact of AI? And why should Switzerland play a key role in this development? Thomas Estier, former researcher of field and space robotics, former deputy vice-president for innovation at EPFL and co-founder of two fast-growing start-ups, offers a pragmatic point of view on trends in robotisation.
Founding ROVENSO in 2016 had a clear objective: using robots to get workers out of danger zones, especially in notoriously risky outdoor industries like oil and gas, chemical mining, waste management, power plants and construction sites. This is considered the third wave of robotisation, and it has only recently started to show results.
3 waves of industrial robotisation
The second wave only recently grew in prominence with the acquisition of Kiva System by Amazon in March 2012. The e-commerce giant was one of the first major players to understand that equipping warehouses with robots was the logical next step in this field: using mobile autonomous robots to move goods within warehouses and redesigning warehouses for robots by installing flexible and moving shelving systems.
Few people know that startups like BlueBotics in Switzerland were already providing such autonomous solutions since 2001. It took 10 further years to get investors onboard and fundamentally change distribution and work processes through robotisation. In fact, the Kiva acquisition initiated an investment hype in robotics that is reaching its peak only today. From 2012, Google started to acquire start-ups in robotics, including the well-known company Boston Dynamics in 2014. Though it wasn’t able to replicate its visionary success in software to hardware, and eventually sold a number of companies (including Boston Dynamics to SoftBank), the fear of missing out on opportunities drove a lot of cash into robotic startups all around the world.
In the field of on-site security monitoring with the help of autonomous robots, which is now the focus area for ROVENSO, three Californian startups alone raised over 60 million USD. On average, this is ten times what European robotics startups expect to raise at an equivalent stage of development. So, how can a Swiss startup with a burn rate of 100’000 CHF per month compete with its US counterparts that burn 1’000’000 CHF in the same period?
Switzerland’s key advantage in robotics
The robotics community, both in the academic and economics fields, knows that Switzerland has one key advantage: the highest density of robotics talent in the world. Smaller than an average Chinese city, Switzerland still has over 25 academic laboratories in robotics, and most of them rank in the top 5% in their respective fields. In addition, the country has 73 startups working in robotics (and 167 in AI and machine learning) and Amazon, Apple, Bosch, Disney, Google, Go-Pro, Facebook, IBM, Magic Leap, Microsoft, Oracle and Samsung are all conducting research in Switzerland.
One reasons for this unparalleled concentration of talent is that Swiss engineers, regardless the university, have a multidisciplinary approach to working. Their knowledge integrates well into lean development teams that the most efficient start-ups favor. Case in point: back in 2016, the two ROVENSO co-founders Beat Geissmann and Lucian Cucu were part of the start-up accelerator HAX in Shenzhen. In China, they were able to build a second-generation robot from scratch over the course of three months. Performing simulations, drawings, programming, electronics, assembling parts and welding cables. Their multidisciplinary experience built the foundation for a strong and agile core team that could flexibly be completed with specialists.
From factories to the field
After securing talent and funding, three more things are needed to get robots out of the factories and into the field: mobility, autonomy and a useful task to fulfill.
Mobility describes the capability of a robot to navigate any terrain that it needs to perform its mission. This is more challenging in complex outdoor environments than on the factory floor. Even a simple urban environment containing steps, holes and stairs is usually deadly for robots. To solve this mobility issue, several strategies are available: more sensors, actuators and intelligence can be added to the robot. This is the path favoured by ANYbotics for its walking robot ANYmal, a spin-off of the ETHZ lab by Marco Hutter. This comes at a cost, in energy consumption and complexity, and a higher price point, but the robot offers the mobility of a donkey, which is a serious benchmark.
For ROVéo, we chose another strategy by building a wheeled robot, the most energy efficient locomotion system in urban environments. We designed a chassis inherited from space rovers developed by EPFL. The robot can climb vertical steps bigger than its wheel size and even bigger than its own ground clearance. It can climb stairs without previously analyzing the environment, as gravity makes the chassis adapt passively without any other actuators except for motorized wheels. We solved a complex robotics problem with simple mechanical means.
The second challenge is autonomy, which means two things: energy- and motion-autonomy. With modern Li-ion batteries, providing energy-autonomy becomes easy. It is technically possible to reach eight hours of continuous operations, which covers a majority of industry use cases.
Unsupervised motion, on the other hand, is a more complex challenge. A robot needs to autonomously locate itself and determine the most efficient route to its destination, while avoiding unexpected obstacles. The first impulse would be to rely on GPS, but the technology is not ideal for two reasons: firstly, in urban environments, some areas close to tall buildings cannot reliably be covered by GPS and secondly, GPS cannot foresee unexpected obstacles. Modelling the outdoors in 2D is not feasible, so you need to build accurate 3D models by using distance measurement lasers, so-called lidars. It is worth noting that some companies claim to use artificial intelligence to solve the challenge of movement-autonomy, when in fact, statistical methods are more reliable and most often used.
Finally, it is important to find an outdoor robotics mission with clear value for potential customers. A critical question for any startup. At ROVENSO, we focus on the security surveillance market. We develop agile robots that perform security and safety monitoring of industrial sites. They patrol autonomously inside and outside buildings, detecting potential intruders and thefts. By using thermography and acoustic analysis, they are also able to safely detect anomalies like fires, liquid and gas leaks that could put people’s lives at risk. This is where artificial intelligence brings most value: by applying a combination of supervised and unsupervised machine learning to data from lidars, night vision, thermal analysis and acoustic analysis. This enables the detection of the three biggest risks to the integrity of assets and stored goods in all industries: intrusion, water damage and fire.
To make a long story short: any industrial application that has a high immobilization cost of assets is an excellent candidate for 24/7 robotized monitoring. The technology is constantly being refined and its possibilities steadily extended. Robotics can help make a difference: by taking humans out of danger zones.