Table of contents

• Improve HMI for ride-hailing efficiency in D1
• Context & main challenge
• Finding the right problem
• A systematic approach
• Improve real efficiency
• Improve perceived efficiency
• Making decisions for ride-hailing
• Impact and deficits

Improve HMI for ride-hailing efficiency in D1

I joined the project in 2022 as a UX designer, collaborating closely with car designers, product managers, and modeling engineers. We focused on foreseeing future use cases and problems after the implementation of new V2X systems, improving the operational efficiency and make EV ride-hailing services more competitive.

Context & main challenge

💡

💡

D1 is the first EV in China designed specifically for ride-hailing. It targets at passengers who value high-end experiences, with the design maximizes back seat comfort, luggage space, and includes a two-way sliding electric door. D1 operates in both B2B and B2C markets. Drivers can rent a D1 from CP (car provider) companies for daily work operations and also use it privately. The mixed business model and use cases create unique design challenges in order to balance operational needs and personal use.

Issues started to arise when D1 tested to operate in Shenzhen, 2021, after it received 10k+ orders. Drivers reported lower daily income compared to gasoline cars, linked to inefficiencies in their work routines after switching to EVs. To combat this, Didi introduced advanced cloud computing technology, connecting vehicles, XiaoJu chargers and phone devices to one network, aiming to streamline operations and improve efficiency.

Finding the right problem

I reviewed the research presentation. The driver group grabbed my attention immediately. They described rough conditions: “I have to work over 15 hours in winter, and the cold hurts my feet”, or, “I slept in the car to work extra hours and got complaints from passengers about the smell”. Charging issues were another big problem: “I have to wait in long lines to charge”, and, “I hesitate to take long-range orders because I worry about running out of battery.” These feedbacks showed how moving to EV had complicated driver’s time management. The added charging needs disrupted the route planning, causing them to work extra hours for same income. Charging anxiety was a serious drawback and could make some drivers reconsider renting a D1.

Next, I dug into data on screen usage. From graphs, 64% of drivers liked the idea to have in-car screens, but 43% said they could do without it. For those who used in-car screens, most frequent features are: car settings (33%), media (33%), and navigation (26%). In contrast, drivers relied heavily on their phones for information (29%), navigation (29%), and social apps (27%). The combined analysis gave me the idea that in-car screens had limited functionality, pushing drivers to rely heavily on phones. Another data point revealed that drivers almost never used media features when passengers were in the car, highlighting a misalignment in previous design emphasis — the focus on entertainment content didn’t match real-world use.

With these insights, I imagined a possible routine for a full-time D1 driver. They are mostly men between 30 and 50 years old, working six days a week and spending 8-12 hours with their car. Didi’s guidelines requires 20mins break every 4 hrs to prevent drowsy driving, totaling about 2 hrs used for relaxing per day. Considering a EV driving range of 350km, fast charging from 20% to 80% takes 30mins, city speed between 30-40km/h, they need 2-3 charging sessions which takes around 2 hrs. With roughly 80% of their time in the car dedicated to work, drivers can earn about the same as gasoline car drivers, if break and wait times overlap efficiently. While real-world situations vary, this estimation suggests that better planning and scheduling could make a significant difference. There is a potential for efficiency gains through helping drivers to plan drive and charge.

A systematic approach

To tackle the problem, I focused on two main aspects: real efficiency and perceived efficiency. Real efficiency means making tangible improvements in how the system operates. Perceived efficiency, on the other hand, is about how drivers feel and understand improvements.

Improve real efficiency

1. Reducing idle time during work

The system smartly suggests orders within range, ensuring drivers get a steady flow of valuable orders. Didi’s batch matching algorithms would minimize total waiting time and handle order surges during peak traffic.

💡

A recommended order will be highlighted once the driver goes online.

2. Reducing idle time during charge

The system finds the best charging spot for each driver, spreading out charging demand to cut down on wait times. Drivers can reserve a spot, and get the car prepared for fast-charging. Drivers who keep good promises, like arriving on time, could earn higher credit scores or receive compensation.

💡

Drivers can use the charging tab and reserve a spot at in-network charging stations.

Improve perceived efficiency

1. Transparent communication

The system always keep the remaining battery range accurate. It also presents thorough info for what’s available near charging stations, allowing drivers to plan breaks efficiently.

2. Heat maps as a convenient tool

Drivers have easy access to heat maps showing area with high demand, helping them choose locations.

3. Drivers dashboard on HMI

The system allow drivers to operate without a mobile phone.

💡

The heatmap filter visualizes real demand and highlights high-demand areas.

💡

The system takes into account the remaining mileage and hides out-of-range orders.

Making decisions for ride-hailing

The design journey began with simple sketches on paper. During discussions with other designers, we agreed on making charging a separate tab, so it would always be easily accessible.

1. Navbar

We had heated debates over the navbar design, after we finalized its selection of features: car settings, driver tools, apps and charging. The discussion was about whether to include the AC controls in the navbar. We reached the agreement to have both temperature displays and controls on, because Chinese ride-hailing passengers have specific requirements towards the AC. The screen size and its placement defined the upper screen as the most accessible area, so we positioned frequently used features at the top. The charging tab, though frequently used, was placed in a less prominent but memorable spot, encouraging drivers to use muscle memory without needing to look.

2. Home screen

Challenges arose as we worked out the layout, which reflected a establishing of new system logic. The old design used a hub-and-spoke model, where the homepage was a central hub for navigating multiple features. This setup made sense for general relaxation, but wasn’t ideal as it only covered the minor use case in daily operation.

I proposed a mixed model, using both hub-and-spoke and flat navigation to serve different scenarios. The system has one “homepage” for accessing general apps, while another “driver’s homepage” that provides operational information and tools without intermediate steps.

3. Trip information hierarchy

One of the hardest design challenges was deciding how to display information in the smart suggestion cards. Take smart order assignment card as an example, each trip contains current location, pickup location, and drop-off location, along with the times, distances, and remaining battery for each of them. Compared to my first draft, the final design emphasized price and battery usage, grouping time and addresses together and clearly separating the trip stages.

4. Unaccepted orders

Another major debate was about handling unaccepted order suggestions. We observed that drivers often ignore orders they perceive as low value, which could disrupt overall efficiency, putting some passengers to wait longer.

I argued that even with the negative outcome, we should always have an alternative option to guarantee drivers’ control and boost error tolerance. The final design we agreed on included a timing bar attached to the button “accept”. The card showing suggested order would vanish if not accepted in time. Drivers could manually choose from smaller cards showing alternative options, and the unaccepted order would be offered to other drivers in the same batch. The batch matching refreshes every 2mins, thus new suggestions based on new match would appear in no more 2 mins.

5. Navigation

Should the left panel be displayed at all times? When designing the navigation interface, I realized that the left panel, if simultaneously displayed with direction info, cluttered the screen and made navigation details hard to see. The direction info couldn’t be integrated into the the left panel either, as navigation appears after accepting an order. When drivers need to communicate with passengers after starting navigation, left panel and navigation instructions co-exist. The final decision was made based on safety considerations. We decided to disable order selection while navigation was active, and integrate passenger communication inside the navigation window. This forced drivers to end navigation before accessing the left panel, reducing the risk of distraction.

6. Online/offline switch

I suggested a switch mechanism to support independent functioning. Drivers toggle between online and offline to switch between order taking view and profile managing view. This would give them access to previously phone-exclusive features such as tasks, trip histories and daily earnings, without adding extra clicks or distractions. The design also helped to protect privacy by separating work use from general use. The working use is like an independent component that can be turned on and off, covering a wider range of needs with easily implemented changes.

Impact and deficits

This design was internally reviewed and tested within a very small scope, using an iPad in a cabin prototype. It demonstrated improvements by reducing 1 to 4 clicks needed to complete general operation-related tasks.

The heat map feature received the most appreciation. The new heat map can be viewed on the center HMI—compared to it being exclusively accessible on the phone app, and the feature is easy to find—compared to drivers complaining it was hard to find or that they didn't even know it existed because it was buried in an unrelated menu.

The smart recommendation is also well-regarded because it saves decision-making energy. However, to evaluate its effectiveness in communicating information, it still needs review from real driving applications, rather than the simulated stationary setting we had when testing.

Time limitations and the conceptual nature of this project also impacted evaluation in accessibility. Major information displays passed the contrast test, but some sub and minor descriptions did not. This suggested room for improvement in future iterations, to meet the standards for in-car displays.

Overall, the design received positive feedback from higher managers. They liked the aesthetic style. The robotic and futuristic tech look has proven to be attractive to male drivers.