Periodic Reporting for period 1 - Safe and Sound Drive (Design of a serious game for cars to help increase driver skills and lower fuel consumption)
Reporting period: 2015-09-01 to 2016-08-31
Although the demonstrated effects were moderate and the concepts developed in the project generally were not preferred over only visual feedback, the results provide valuable knowledge for further development of interactive sounds for user interfaces in cars in general, and for eco-coaching in particular. The ideas will be further explored in future studies of interactive sounds, and a study of interactive sounds for eco-coaching based on active noise control in trucks is already planned in cooperation between Scania, Luleå University of Technology and Interactive Institute Swedish ICT, financed by the Swedish Energy Agency. Software developed in the Safe and Sound Drive project will be used for further development and evaluation of sound based and multisensory user interfaces in vehicles. Contacts between Luleå University of Technology and Anglia Ruskin University are well established, and further collaboration is planned.
The results were disseminated through participation in conferences, publication of conference papers, demonstrations at two science festivals (Cambridge Festival of Ideas 2015 and Cambridge Science Festival 2016), a demonstration at school event (My Smarter Essex, Chelmsford 2016), workshops with Volvo Car Corporation and discussions with Scania trucks. Outcomes of the project have also been disseminated through regular blog, Twitter and Facebook posts. The demonstration software can and will be used for future demonstrations at science festivals, science centres and other popular science events. Parts of the methods developed in the project are already implemented in Volvo Car Corporation’s design procedures for interactive sounds and will be implemented at Scania through the planned continuation project.
Objective 1: Design a demonstrator of a serious game supporting the driver to increase driver skills, lower fuel consumption and allowing safe and environmentally friendly competition with friends.
Objective 2: Develop and release a mobile application version of the game for iOS, available for download from App Store (this will at a later stage be implemented also for Android and Windows).
Objective 3: Measure how the designed game affects fuel consumption, speed and safety.
Objective 4: Training in computer gaming art and technology.
Objective 5: Training in creative use of sound design.
Objective 6: Digital dissemination training – maximising impact of research.
Objective 7: Training in mobile application programming and technology.
Objective 8: Training in commercialisation of mobile applications.
Objective 9: Public dissemination by demonstrations at science centres in the UK and Sweden, and presentations and demonstrations at festivals and events hosted by Anglia Ruskin University and its networks.
The project followed a user-centred research through design scheme. A demonstrator of an audio-only serious game supporting the driver to lower fuel consumption was design. Volvo Car Corporation supported the project with access to a car, full access to sensors and systems in the car. The goal was to develop a mobile applications version of the game for iOS. Because of practical reasons, mainly Nykänen’s previous knowledge in development for the Windows and Android platforms, and because of new knowledge in game engine technology and audio middleware for games gained during the project, the demonstrator was developed using Unreal Engine. The demonstrator was implemented for Windows and demonstrated in a driving simulator and in a car. Future work includes implementation for Android and Apple platforms for public download.
User-Centred Agile Methods
Design of user interfaces is a multidisciplinary task, requiring deep knowledge in psychophysics (or in this particular case, psychoacoustics) and psychology, as well as methods and procedures for design and product development. Knowledge from psychoacoustics and psychology aids the development process, and enables early decisions on potentially successful design routes, avoiding known pitfalls. Examples in the Safe and Sound Drive case are the choice of modality (hearing over vision), choice of frequency content of the sounds to avoid masking, spatial positioning of sound sources to help the driver to keep attention on the road, awareness of the risk of cognitively overloading the driver with a too complex task, etc. However, the design options are still vast, and knowledge is built along the path of the project. A good final design is crucially dependent on how this knowledge is built and taken care of all the way through. A user-centred agile approach was used. In particular, agile game development was explored, as game design is a field where both graphics and sounds are essential parts of the product. The core idea is to organise the work into short iterations (sprints), listening to the evolving product frequently and repeatedly throughout the design process, and get users into the loop early on. Game engine technology was used for rapid prototyping of concepts.
The Sound Design and Research Process
A user-centred agile method was applied to the Safe and Sound Drive project, including rapid and numerous iterations, and supported by the use of a large number of sound sketches. The process was inspired by Scrum, and major elements from the Scrum framework were used, for example: the division of work into time-boxed sprints; User Stories, Product Backlog and Sprint backlogs; prioritisation of backlog items; and the general aim for always delivering a working result from a sprint. Repeatedly during process, qualitative studies and analyses were performed based on diary notes by the participating designers, Contextual Inquiry Interviews, and observation of participants. This approach increases the chances of success in identifying a design that significantly affects fuel consumption and driver behaviour. Finally, the design was evaluated in a simulator study where driver behaviour was measured.
In the start-up/release planning meeting a number of user stories were determined. In agile methods, user stories are typically one sentence descriptions of what a user wants to do with the product, written on the form: “As [kind of user] I would like [a feature] so that I can [reach a goal]”. The following user stories were defined in the start-up/release planning phase:
(i) As a driver I would like to hear when the fuel consumption is low.
(ii) As a driver I would like to hear when the fuel consumption is high.
(iii) As a driver I would like to hear when I am driving too fast or too slow.
(iv) As a driver I would like to hear when I am using too high or too low torque (too much or too little throttle).
(v) As a driver I would like to hear if I am driving at an even or fluctuating speed.
The available input from the car to the system was: speed, torque, speed limit, acceleration, and rate of change of torque. At the time of the project start, the car did not deliver momentary fuel consumption. Therefore, the two first user stories were not implemented. In Sprints 3 to 5 the design work was focussed on sounds to encourage an even speed as this is a key factor for fuel efficient driving and easily measurable in lab settings.
Design Sprint 1 – Prototype implemented in racing game and in car
The delivery from Design Sprint 1 was a fully working prototype, first implemented in a computer racing game and then in a real car. The objective was to allow the designers to develop and test-drive a number of sound concepts. It was assumed that experiencing the sounds in the right context is essential for making the right decisions.
Design Sprint 2 – Contextual Inquiry Interviews
The planned outputs from Sprint 2 were prototypes of sound concepts working flawlessly with all technical and perceptual issues raised in the review of Design Sprint 1 corrected. Nine concepts were developed and assessed in Contextual Inquiry interviews with three drivers in a real car on a test track. These interviews showed that:
- Opinions on beeps, and tones varied among the participants.
- Speech messages were generally not appreciated. The human nature of the voice messages may give rise to expectations on the system of performing and reacting like a human.
- Concepts based on music or podcast contents were generally well received. Alteration of music and podcasts, for example by adjusting BPM, music mix, volume or spectral content, could be used for providing drivers with continuous information, with a high level of acceptability.
The designers’ reflections can be summarised into the following points:
- Immediate responses to driver actions were generally perceived as logical and desirable.
- Individualisation and possibility to switch the system on and off are probably necessary for acceptance, as preferences for sounds and music varies. Altering a play list or radio program chosen by the driver and elaborating the content could be a way of dealing with the large differences in preferences.
- Addition of sounds, for example drums, when approaching the target speed created a feeling of flow. Therefore, concepts based on composition of sounds enhancing the feeling of flow at the target speed were developed.
- Simulated driving in the designers’ studios was useful for experiencing the interactive sounds. However, the experience varied much between the studio and the real car. It was crucial to do final adjustments, e.g. adjusting levels, equalisation, responses to events, etc., in the car while driving.
Design Sprint 3 – Game Demonstrated at Cambridge Festival of Ideas 2015
Gaming elements were introduced into the racing game/simulator. The game was intended for testing design concepts through demonstrations to larger audiences at science festivals. The game prototype included four sound concepts developed from the most successful designs identified in Sprint 2:
1. Speech only podcast. Treble or bass were gradually decreased as the deviation from the target speed increased, treble when speed was too low and bass when speed was too high. A buzzing sound was played when speed started to deviate from the target speed, and three dings indicated that the target speed had been approached.
2. Music, free choice from Spotify. Treble and bass were adjusted as in Concept 1, and the same buzzing sound and dings were used.
3. Loop based music. Composed for the project. Three states were used: too high, too low and correct speed. The character of the music and the lyrics changed with state.
4. Music mix. Circus by Daisy and the Dark (from the Red Planet album app released for iOS at the App Store, 2015). Remixed depending on speed. Drums removed when deviating from target speed. At too high speed the mix was made intense and at too low speed calm.
The interactive sounds provided support to maintain a constant speed. Measures of simulated fuel consumption and integrated deviation from a given target speed were used to calculate an “eco-score”. The game prototype was based on an open source racing simulator (CORS based on TORCS). OBD-II messages were sent from the racing simulator to a Pure Data patch that interpreted them and controlled software synthesizers in Ableton Live through MIDI.
The prototype was launched at Cambridge Festival of Ideas 2015, where approximately 100 participants tried the game. Feedback to the project was based on one of the designers observing and interacting with the participants. It was felt that the opportunity to interact with such a large number of people gave a deeper understanding of user behaviour in general, and advantages and drawbacks with the design concepts in particular. This experience is supported by literature showing that having the development team taking part in user testing increases their understanding of the usability problems and makes them more sensitive to problems encountered by users.
Design Sprint 4 – Game Demonstrated at Cambridge Science Festival 2016
The four concepts developed in Sprint 1-3 were implemented in a driving game developed using the Unreal Engine 4 game engine. The FMOD game audio middleware was used to allow rapid prototyping of the interactive sounds. The main reason for this change of development platform was to allow for more flexibility in the design of scenarios and game ideas, as well as easy integration with game audio middleware. The new version was demonstrated at Cambridge Science Festival 2016, where more than 300 participants completed a test drive and eco-scores were calculated and presented. A very simple environment and car asset were used (see Figure 3), giving the impression of work in progress and making participants more aware of the objective of the demonstration, helping them pay less attention to details in the game play that were not under investigation.
Design Sprint 5 – Experiment Based on Simulated Driving
Three concepts were developed based on concepts from Sprint 4. In addition, a baseline condition with only simulated interior car sound was included. A driving scenario suitable for a lab experiment was designed. It consisted of a road resembling a rural road without other traffic, with 30 mph and 60 mph speed limit sections alternating during the test drive. The participants were asked to try to drive as close to the speed limit as possible. The driving task was intentionally made quite difficult, with a car simulation that was sensitive to hills and to mistakes in the control of the vehicle. The car asset visible in the game used in Sprint 4 was hidden and a plain graphical presentation of speed and gear was added. 16 subjects participated, 7 female and 9 male, mean age 28 years (SD 14 years), all with a valid driving licence. The experiment started with a training session where all four conditions were driven for approximately 3.5 minutes each.
After the training session the test session started. In the test session the four conditions were driven on the same track, but the track was driven two laps (i.e. approx. 7 min. per condition). After having finished driving one condition the participants were asked to rate how they felt while driving using the 9-point Self-Assessment Manikin (SAM) scales for measuring pleasure, arousal and dominance. In addition, average speed and integrated deviation from the speed limit (target speed) were measured.
Finally, the participants were given the opportunity to give their opinions on the different driver assistance systems. This was made by first driving each system for a short distance, just to remind them about the system. They could abort driving at any time when they felt ready to indicate their preferences and give comments on an assessment sheet. Assessments were made on two 9-point scales: “How did you like this driver assistance?” on a scale ranging from “Disliked very much” to “Liked very much”, and: “Would you use this driver assistance system if it was installed in your car?” on a scale ranging from “Not likely at all” to “Very likely”. In addition, an open box for free comments was provided.
A significant difference in average speed between the baseline condition (no audio-based driver assistance) and the condition with treble and bass adjustments of music was found. Participants drove in average 1.1±1.0 mph slower when the music based driver assistance system was used compared to the baseline condition (p <.05). No other speed related measures differed significantly between the conditions. The results suggest that this kind of continuous audio-based assistance can have some effect on people’s ability to keep an even and suitable speed during a short drive. A likely explanation for the limited effect is that visual information dominated the determination of the participants’ behaviour and performance in this experiment.
Further, a significant difference in estimated arousal was found between the baseline condition (no audio-based driver assistance) and the condition with treble and bass adjustments of music. Participants were less excited/more relaxed when the music based driver assistance system was used compared to the baseline condition (p < .05). No other feelings were significantly different between the conditions.
Significant differences in the likelihood for a subject to use the systems in his/her own car were found (p<.05). Post-hoc tests showed that participants rated the likelihood that they would use the system based on music lower than the baseline condition (p<.05), and the system based on podcasts lower than the baseline condition. It is apparent from the results that the participants preferred driving without the audio based driver assistance systems, despite the positive effect the systems had on driver behaviour. It also indicated that the alteration of music mix was more likely to be used in real driving than th
e other two