AViD.
How might police interact with driverless trucks?
How Might Police Interact with Autonomous Trucks?
In partnership with Torc Robotics, we aim to address the question: How might police interact with driverless trucks? Over the course of eight months, we worked diligently to identify the most impactful and feasible interactions between police officers and autonomous trucks. Our project focused on designing a tool to streamline truck inspections, making the process faster and more efficient.
We developed an intuitive interface that incorporates both voice and touch controls to enhance safety, reliability, and ease of use. This innovation helps officers complete inspections quickly while ensuring that trucks can return to the road sooner.
“Team Construcktive” was made up of Sohail Bagheri, Roshan Selladurai, Pradeep Vegireddi, Vera Wei, and myself.
Link to full project website (Opens in a new window)
Solution.
AViD: The Autonomous Vehicle Inspection Device
Meet AViD (Autonomous Vehicle Interaction Device)—a testament to our avid efforts in designing usable solutions for law enforcement to communicate with Torc’s driverless trucks. AViD is a mobile application that includes unique features such as multimodal input, automatic mode for inspecting the vehicle, and autofill for inspection reports.
Key features of the application design include:
Simple two-factor authentication
Automatic detection and sorting of nearby vehicles
Real-time display of carrier and vehicle status
Ability to select inspection levels
Electronic documents for viewing and exporting information
Voice input for hands-free operation
Testing of inspection items typically operated by drivers
Progress tracking for inspection steps
Exporting inspection data to reports
Problem.
Addressing Long Haul Challenges with Autonomous Trucks
Our economy depends very heavily on trucks, whether it's the food we eat or the device you're viewing this website on. At the same time, however, qualified truck drivers are leaving the industry in search of jobs that offer better pay, benefits, and working conditions. This leads to a growing mismatch between demand and supply.
“70% of the tonnage that moves across America depends on trucks”
“80,000 driver roles in America were left unfilled in 2021—a record shortage”
Currently, when trucks transport goods from Point A to Point B, their journey would roughly look like the diagram below, where there are three separate segments, each handled by a different driver. The middle mile is usually the longest segment. The long haul drivers who navigate the middle mile drive upwards of 11 hours a day on the highway. These long hours, combined with the monotony of highway driving, often result in dangerous levels of fatigue and distraction. In addition, long haul drivers are often subject to unhealthy lifestyles (e.g., sleep in truck, lack of nutritious food) and spend most of their time away from home.
Recognizing the long haul truck drivers' plights and the pressure that driver shortages put on businesses, Torc (partnered with Daimler Truck AG) is working to automate the middle mile. By introducing autonomous trucks to the middle mile, Torc helps to reallocate scarce driver resources to where they are needed the most: the first and last mile. Skilled drivers are especially valuable at the first and last mile, as those routes often include local roads and urban environments with a higher density of people and objects. By contrast, the middle mile mostly consists of highways, which are often fixed and predictable in a way that's easier for self-driving vehicles to navigate.
Ultimately, Torc's strategic focus on the middle mile will benefit multiple stakeholder groups
Long-haul truck drivers will be able to shift from the distressing middle-mile lifestyle to first and last mile routes, thus improving their quality of life and job satisfaction.
Businesses + retailers will experience lower supply chain costs, as autonomous trucks introduce efficiency to the longest and most costly segment of the journey.
End customers will get faster deliveries, as autonomous trucks complete the middle mile in less time than human drivers.
It seems like Torc's autonomous trucks are bound to make everyone happy, but why haven't they become the norm? In reality, we won't be seeing them at mass scale until those who regulate road conditions (e.g., police officers) can communicate with them. In other words, without the ability to smoothly interact with autonomous trucks and assess their compliance to safety standards, law enforcement will not be able to trust their operation.
Even when autonomous trucks do become the norm and police have a way to interact with them, the means of interaction shouldn't come without strategy. When police decide to pull over an autonomous truck, the truck will need to stop; the longer the truck stays halted, the more likely it will fail to deliver on time and lead to additional costs for businesses. Effective and efficient communication is key here.
Altogether, this begs the question...
How might we streamline interactions between police and autonomous trucks?
Process.
Keeping the question of “How might we streamline interactions between police and autonomous trucks?" in mind, we defined some key research goals:
Identify how and why police would pull over or interact with a commercial truck today
Explore and understand possible interactions between police and autonomous vehicles, assuming they have not interacted prior
Learn about needs, potential pain points, and barriers associated with communication between police and commercial trucks
Investigate information transfer and communication systems within the autonomous vehicle sector and analogous domains
As we gathered more information, we found that there are many reasons for police to approach a truck, but the likelihood of the truck being able to continue its trip is not equal across all contexts of interaction. For example, if the truck is involved in a road collision and gets damaged, any efforts put towards streamlining police-truck interaction would likely be futile, as the truck will still not be able to return to the road due to mechanical breakdowns.
For the scope of this project, we’re looking at cases where the truck can still feasibly return to the road, and, among those scenarios, we’re in search for the most frequent and most time-consuming interactions. Time is money, so fast-tracking these interactions would generate the most added value for Torc and its customers. We also want to focus on interactions that are predictable to some extent, as higher predictability would more likely pave way for the automation of interaction without involving external assistance (e.g., dispatchers or Mission Control).
Optimizing Police-Truck Interactions: Focus on Inspections
We ultimately found truck inspection to be the most promising type of police-truck interaction that, if streamlined and expedited, will help maximize value for Torc and other stakeholders. Inspections occur when the police need to verify whether the truck is in an acceptable condition to safely operate on the road. Compared to other interactions such as those prompted by illegal activity or road accidents, inspections happen more consistently and don't diminish the truck's chances of returning to the road, so long as the vehicle meets safety standards.
To fulfill our goals, we employed a variety of research methods, making sure to build a holistic understanding of the problem space and hear from different perspectives.
20+articles and documents to better understand the problem domain and investigate existing research and tools in the space
22 hours of interviews to surface different stakeholders' detailed thoughts, feelings, and beliefs
6 in-field visits to observe our target users' current processes and practices in an authentic, natural context
Insights.
Keeping this in mind, we are prioritizing truck inspections as the focus of our efforts. The procedures involved in inspections can include anything from verifying documents to examining the vehicle’s interior and exterior. If we can automate inspection, it's likely that we have found a way to automate all types of necessary interactions between police and autonomous trucks.
Through our research on police-truck interactions (with a focus on truck inspections), we gathered a wealth of information and ultimately extrapolated four key insights. These insights are important to consider because they hold implications for how a solution that best streamlines interactions between police and autonomous trucks might look.
-
The more experience police officers have with inspecting trucks, the more they rely on indicators such as vehicle size, reputation, and safety rating to make systematic decisions about which trucks to inspect.
-
Although there is a standard checklist of what to do during inspections, a police officer's approach is non-linear, contextual, and dependent on specific instructions from their department or the carrier.
-
Police officers currently rely on verbal and gestural language to communicate with truck drivers. They worry that, when interacting with driverless trucks, their experience will no longer be relevant and communication might not be as fluid or effective.
-
The lack of standardization in the technology infrastructure for law enforcement and trucks, combined with highly variable environmental constraints, eliminates the ability to assume any consistency across different interaction contexts.
Design.
There’s no doubt that many different possibilities exist for enabling and streamlining interactions between officers and autonomous trucks. Before identifying and zeroing in on the most viable and valuable opportunities, we let our minds run wild and untethered to explore the breadth of options. From device types to input and output modalities... We considered many different permutations before slimming the pool of ideas.
We generated ideas primarily based on the following principles:
Reliability: Officers must feel their safety and privacy are secure to prevent friction, especially given the unfamiliarity of driverless vehicles.
Learnability: The tool should align with existing inspection software patterns, ensuring an intuitive experience with minimal training.
Efficiency: Quick, effective inspections allow for faster return of trucks to the road, boosting profitability for Torc.
We discussed and tested ideas and prototypes with officers from across the country, as we anticipated regional differences to create discrepancies among our target users' experiences.
Final design.
Common & simple two-factor authentication
After testing with different variants of authentication such as entering a badge number and taking a photo of the police vehicle’s license plate, etc., we ultimately opted for a simple and common two-factor authentication flow, which already exists for most software applications that officers currently use.
Display basic carrier and vehicle status information at all times
To minimize the risk of error (e.g., inspecting a vehicle that isn’t the one the officer’s device is connected to) and danger (e.g., standing in front of the vehicle while its brakes are released), officers want to see identification and status information about the connected vehicle at all times. This information is available after the officer is accounted
Automatic detection and sorting of nearby vehicles
For the purposes of helping officers identify and connect to vehicles, we leverage near-field communication (NFC) between the vehicle and the officer’s device, an entry point that spares officers the manual effort of finding and scanning a code. Officers mentioned that they may want to sort vehicles differently depending on the context. Since they were commonly mentioned, we include time of last inspection and distance from the officer as the two sorting parameters in AViD.
Allow officers to select a specific level of inspection
Across different inspection levels, there are a variety of tasks, including checking driver credentials and inspecting the actual vehicle. However, not all tasks will be relevant for each level. For instance, the Level V Vehicle-Only Inspection naturally won’t involve checking information related to a driver. Selectively showing or hiding certain action items based on the intended inspection level seems to spare officers the cognitive load of deciding which actions are relevant.
Provide electronic documents to view and export info from
Furthermore, officers voiced the desire to be able to export basic information from shipping documents to the inspection report they need to complete. As of now, they typically need to input the carrier information manually. However, this can be time-consuming and increase the opportunity for human error. Any effort to decrease the need for manual input for inspection report seems welcome.
View all inspection items and select one to begin inspecting
Different officers may have varying preferences for how they process information, so we organized inspection items and categories into two views: list view and truck view. The list view shows a scrollable list of all the inspection item categories, each of which can be expanded to show more granular items. The truck view doesn’t show the categories all at once, but instead shows different regions of the truck in a visual format and includes the inspection item categories as hotspots on corresponding regions.
Marking the status of each item and adding optional notes
Currently, with an unformatted notepad, officers lack a systematic way to document information while inspecting the vehicle. With AViD, officers can easily and consistently check off inspection steps. We hypothesized that this would help reduce the officer’s cognitive load, and many officers were quick to validate this impact. Furthermore, they can optionally add a more descriptive note with images, voice memos, etc. to elaborate on their decision.
Exporting inspection data to inspection report
Currently, officers need to complete a mandatory inspection report after each inspection. Requiring officers to do the same tasks on both the inspection report and AViD would be redundant work. As such, we envision an option to export data directly to the officer’s inspection report. The officer can choose what information they want to include or exclude from the export.
Voice input as an alternative to touching the screen
We got early feedback that touchscreen interactions are not always the most feasible. For example, officers may get their hands greasy or wet during inspections or wear gloves in cold weather. In these circumstances, it would be critical to provide an alternative method of user input. Voice input proved to be one of the most well-received ideas among officers we spoke to.
Testing inspection items that drivers would usually operate
Because the task of activating inspection items such as headlamps and brakes is a repetitive task throughout the inspection, we wanted to make sure the mechanism for doing so stayed at the most easily accessible part of the screen: the bottom. We also included an “automatic” mode that can be toggled on or off at any point. When turned on, activate inspection items one by one automatically instead of manually requesting activation each time
Seeing the progress in completion of inspection steps
Because the inspection process can be long and exhaustive, officers value being able to see what they have and have not completed. This can help them easily recognize where they are in the process and minimize their margin for error. Indicators of progress are present in the list view, truck view, and a persistent progress bar at the top of the screen.
