Automation for the people: opportunities and challenges of humanitarian robotics

On a crisp late-September morning in Panga, Nepal, just outside Kathmandu, a small aerial robot, the DJI Phantom 3, floats through a jagged landscape of damaged buildings and uncleared rubble. A high-resolution camera affixed to its underbelly silently siphons up image after image. Onboard sensors stabilise and geolocate both the robot and the pictures, allowing its data collection mission to follow a precise predefined pathway set only minutes before in a smartphone application. Later that afternoon a group of software engineers and students from Kathmandu University assemble the data into orthorectified mosaic maps Orthorectification means the creation of a photographic map which shows locations in their accurate spatial positions by means of the removal of various aspects of image distortion. For more information see https://trac.osgeo.org/ossim/wiki/orthorectification. which can be draped over digital models of the earth for the sake of analysis and interpretation. The very same data contains point cloud measurements A ‘point cloud’ is a collection of data points generated in three-dimensional space, which can be processed into a model of objects located within that space. On the use of UAVs to generate 3D point clouds, see A. Ansar, ‘Use of Point Cloud with Low-cost System for 3D Mapping’, http://ieeexplore.ieee.org. of the distance between the camera and ground features, allowing for the creation of three-dimensional models to assess damaged infrastructure. The next morning a team of Nepali citizens and scientists, along with international technology professionals and aid workers, all convened by UAViators (the Humanitarian UAV Network), examines the maps and models in search of new, timely and more detailed perspectives than were previously thought possible on rebuilding in the wake of the April 2015 earthquake.

Aerial robots are the first wave of robotics to impact the humanitarian space. Prior to the broad-based introduction of aerial robotics in humanitarian assistance, ground robots, or ‘unmanned ground vehicles’ (UGVs), were used in mine clearance and search and rescue. In neither case, though, did these applications spark a broad wave of technical innovation in the humanitarian sector. See Centre for Robot-Assisted Search and Rescue: http://crasar.org. They will certainly not be the last. Popularly known as Unmanned Aerial Vehicles (UAVs) or ‘drones’, aerial robots have already been used many times in many different situations to collect data in support of disaster response and recovery efforts. The International Organisation for Migration (IOM) has been using aerial robots in Haiti since 2012 to capture aerial imagery to assess disaster damage and displacement. In 2013, Medair and Catholic Relief Services (CRS) used aerial robots to collect imagery to inform their reconstruction and rebuilding efforts following Typhoon Haiyan in the Philippines. The following year, Médecins Sans Frontières (MSF) and the World Health Organisation (WHO) piloted the use of aerial robots for the delivery of small medical payloads (vaccines, medicines) in Papua New Guinea and Bhutan respectively. In 2015, the World Bank used aerial robots in Vanuatu and Tanzania to support disaster response and risk reduction efforts. Several agencies used aerial robots for search and rescue, situational awareness and mapping following the earthquakes in Nepal in 2015. In 2016, the UN Children’s Fund (UNICEF) will pilot the use of aerial robots for medical payload transportation in Malawi. This year Redline will also launch Rwanda’s first Droneport network to facilitate routine long distance cargo delivery using aerial robots.

Sensing and acting at a distance

The common denominator in all of these efforts is the capability of robotics technologies, aerial and otherwise, to sense and act at a distance at reasonable cost with varying degrees of autonomy and intelligence. Rather than send a human up in an aircraft to take photographs of a flood zone, refugee camp or agricultural field, we can now send an aerial robot to do the job at far lower cost and higher data quality. In the time it takes to send a human on a motorbike to retrieve and deliver laboratory samples over muddy and sometimes impassable roads, we can now potentially send a small drone with secure cargo capacity back and forth multiple times, speeding up entire public health diagnostic systems. Rather than sending people in boats into urban flood zones to measure water contaminants we can now send small fleets of semi-autonomous marine robots to gather and analyse data more safely and quickly over a much wider area. Rather than risk sending people into minefields to determine optimal patterns of explosives removal we can now send rugged ground robots.

In each of these cases we can begin to detect not only new opportunities to do what disaster relief professionals and humanitarian agencies have always done, only faster, safer, cheaper, more efficiently or more accurately; we can also detect the outlines of possible new types of dynamic, flexible and adaptive public service and humanitarian systems. Medical payload delivery is perhaps the clearest case, even if many of the core technologies have not been entirely proven. Failure of the drug supply chain, particularly at the last mile in remote areas with low settlement density, patchy health infrastructure and poor transport systems, is one of the leading causes of serious health problems including maternal mortality, childhood pneumonia and diarrheal disease. David Jamieson, ‘The Health Supply Chain: Still the Cinderella of Development?’, 18 July 2015, http://www.theguardian.com. Failures due to poor transport systems are compounded by weak centralised procurement and distribution systems, which lack timely information on changes in local demand and the capacity to respond to quickly changing information. In the most obvious sense drone delivery systems could avoid the problem of road transport conditions altogether, alleviating a key logistical blockage. But that important detail looks almost minor compared to the ways that drone delivery could alter core attributes of health systems.

Imagine, for instance, that trained community health workers spread throughout small villages could determine specific local demand for basic needs like vaccines, nutritional supplements and antibiotics, and communicate that demand to regional distribution centres via SMS or cellular data channels. Rather than directing medical supplies through health facilities which may not be sufficiently responsive, they could request supplies directly to remote points of care, such as household vaccination campaigns or mobile clinics. During the Ebola crisis in West Africa, we know that cases of preventable childhood illness spiked in part because health facilities in Liberia, Sierra Leone and Guinea had become sites of disease transmission and were therefore avoided by much of the population. Effective drone delivery systems tied to strong community health worker programmes might have circumvented this problem.

Social automation and social collaboration

Perhaps it goes without saying, but it’s nevertheless worth remembering that, no matter how promising or how well designed they are, robots will not accomplish significant humanitarian gains on their own. The opportunities of social automation for social good are inevitably tied to the challenges of improving human–machine collaboration within the context of integrating robotics into humanitarian systems through a combination of standards, evidence and institutions.

The concept of improved social collaboration with robotics technologies goes back to the foundations of the information age and J. C. R. Licklider’s depiction of ‘man-computer symbiosis’. J. C. R. Licklider, Man-Computer Symbiosis, IRE Transactions on Human Factors in Electronics, March 1960.  As Licklider framed it in 1960, effective social automation systems depend on a viable division of labour between human-centric goal orientation and machine-centric task performance: ‘In the anticipated symbiotic partnership, men [sic] will set the goals, formulate the hypotheses, determine the criteria, and perform the evaluations. Computing machines will do the routinizable work that must be done to prepare the way for insights and decisions in technical and scientific thinking’. A shifting but nevertheless persistent boundary exists between what people and computers, of which robots are a variant, can each do well. If that boundary is managed intelligently and creatively, the resulting socio- technical ensembles can produce social gains which neither humans nor computers could achieve on their own. Framed in this way, the real question for humanitarians is not whether they ought to use robotics at all, or whether robotics ought somehow to determine the shape of our public goods, but rather what kinds of possibilities are achievable through their combination with social good efforts, and what kinds of new institutional arrangements will be required to achieve those possibilities.

As a first step, the humanitarian community should develop an international code of conduct on the use of robotics in humanitarian aid. This could usefully draw on the International Code of Conduct on Humanitarian UAVs developed by the Humanitarian UAV Network (UAViators) in close collaboration with dozens of humanitarian organisations. UAViators Humanitarian UAV Code of Conduct and Guidelines: https://docs.google.com/document/. The UAViators Code of Conduct does not proscribe specific behaviours or structures, but sets in place basic standards, such as the requirement that humanitarian benefits outweigh risks to people’s safety, in order to shape the applications of aerial robotics within clear ethical purposes and driven by clear social goals. Adherence to these basic standards is vital to ensure things like regulatory approval by the range of national authorities with sovereign control over airspace, customs and many other legal and policy dimensions of humanitarian operations. The UAViators Code of Conduct is being extended to include data collection and payload transportation. Members of the UAViators network and their partners are in turn working to shape national policies to provide for defined humanitarian exemptions and clear approval processes. Extending the Code of Conduct and this type of policy advocacy further to include terrestrial and maritime robotics would be easier than starting from a blank slate.

Alongside the establishment and implementation of a Code of Conduct for humanitarian robotics, the humanitarian community needs a much better evidence base to provide insight into what constitutes effective robotics applications. Humanitarian benefits need to be proved, not assumed. For instance, although the conceptual and technical basis is clear for developing automated drone-based delivery systems, there is as yet no established cost-effectiveness or health impact analysis to demonstrate the benefits of this approach. While there have been a substantial number of humanitarian drone mapping projects there is still little well-documented evidence, and even less with real methodological rigour, for the effect those projects have had on specific project outcomes. Without the creation of a strong empirical evidence base the robotics field risks missing out on high-quality and high-impact field applications just as the humanitarian community risks falling further behind the exponential growth of new technologies.

Improved standards and evidence for humanitarian robotics also requires new types of localised institutions in order to generate genuinely responsive and effective projects. Humanitarian UAV operators faced myriad challenges in the immediate aftermath of the Nepal earthquake, from regulatory confusion to community suspicion. Several communities were overflown and imaged by different organisations repeatedly without being informed that the flights were taking place or being asked for their consent; data and analysis were also not shared with affected communities and government representatives. UAViators returned to Panga in Nepal in the following months to produce high-quality maps and 3D models, and to create, in conjunction with Kathmandu University, Kathmandu Living Labs, DJI and Pix4D, a new type of localised innovation hub called Kathmandu Flying Labs (KFL). While the immediate goal of KFL is to train local experts and establish the basis for Nepalese-driven demand for aerial robotics, in the future KFL may be able to step into the institutional void which led to so many coordination and communications problems during the post-earthquake response.

From UAViators to WeRobotics

In order to keep up with the rapid pace of global social and technological change, the UAViators network is in the process of transforming into a new institutional framework called WeRobotics. Patrick Meier, ‘Introducing WeRobotics’, 16 November 2015, http://irevolution.net/2015/11/16/introducing-werobotics. Aerial robots may be the first wave of robotics innovation to hit the humanitarian community, but the second and third waves are already on the horizon: industry and academia are making tremendous strides in both terrestrial and maritime robotics like the self-driving vehicles developed by Google and others and the Autonomous Underwater Vehicles (AUVs) being used for environmental research. Like aerial robots, terrestrial and maritime robots will significantly extend people’s ability to collect data and transport payloads in many of the world’s most vulnerable societies. Fortunately we aren’t starting from scratch this time.

WeRobotics represents an effort to apply the lessons learned so far from the use of UAVs for humanitarian aid, global development and environment protection to the fields of robotics and social automation. The heart of this effort is the creation of a global network of innovation hubs modelled on the kind of activity that has proven successful for Kathmandu Flying Labs. Over the next three years, WeRobotics is co-creating globally networked city- level innovation labs with technology and social good partners in cities experiencing cascading risks, rapid development and serious environmental threats, such as Jakarta, Monrovia, Port-au-Prince and Santiago. These ‘Flying Labs’ provide aid, development and environmental organisations with direct access to promising robotics technologies, connecting NGOs, academics and governments with global technology partners and cultivating new locally-owned organisations and trained technology workers. In the process, a new kind of localised but globally networked humanitarianism may emerge, grounded in ethics, dynamically responsive to the needs of the most vulnerable and founded upon strong collaborations with the remarkable new robotics technologies soon to be suffusing our air, land and seas.

Andrew Schroeder and Patrick Meier are co-founders of WeRobotics.