Towards an intelligent planning system

Towards an intelligent planning system

Towards an intelligent planning system S Smith Project management is a highly skilled task which is vital for business success, but current tools prov...

835KB Sizes 4 Downloads 115 Views

Towards an intelligent planning system S Smith Project management is a highly skilled task which is vital for business success, but current tools provide limited support for this critical management function. The paper outlines a proposal for engineering a new tool, based on knowledge-based-system techniques, which will extend the scope of PM support. The proposed tool would be built with conventional planning systems to harness the ,~~~fi,l fm-il;tior nf m,cL tm.Jr n-A tn n.>n;A thn roontoriwn Las, -1, UL“LL‘Cll “J .YucLI‘ ‘““‘J, urcu L” UY”lW W&G rGGr*LcI G”fj of existing plans from scratch. For a company, the bene~ts of such a tool would be significant. It would provide better support for PM by replicating the expertise of good project managers, enforcing company standards, and reducing the managers’ workloads. For a practising manager, the benefits would include an intelligent system which was easy to use, assisted with the druwing up of plans, could nnnl\fID 0*/r LtICLycIC ,.r;t;nrrn ycurw, “L”n a,nd provt?,n advice o,n UISU‘J*GUIIU replanning. Keywords: knowledge-based systems, intelligent planning, project-management support

many 01 me concepts in this paper have been explored by the UK Alvey PLANIT club, which examined the application of expert systems to medium-term activitybased planning’. A key deliverable of PLANIT was a series of concept-proving expert systems. The way is now open to exploit the results of this and related research. This paper takes further the ideas expressed in PLANIT, and outlines a proposal for an intelligent planning system. Project management (PMj is a highiy skiiied task which is vital for business success. However, current PM tools provide very little decision support for critical functions to line and project managers. Most tools provide facilities for drawing networks, and a limited amount of analysis, such as cost modelling and criticalpath analysis (CPA). It is necessary to extend the functionality to address this deficiency. The analysis should cover support across

BMSLtd, 1 Coldbath Square, Farringdon, London ECIR 5HL, UK Vol 10 No 4 November

1992

project objectives, project constraints, l resources, e products, e.g. services or deliverables. 0

*

This means that not only can, say, the critical path be detected9 but also other potential and consequential problems can be highlighted. These can be summarized for senior managers, so that they can focus their attention on the impact on key business parameters. However, it is not sufficient merely to detect problems, especially for a lengthy or complex plan. Decisionsupport facilities for identifying opportunities for recovery are also required. Moreover, as many managers do not find existing systems amenable, there is a need for r.‘---“‘. nlnnnino a ” cvctmnc cz_r! interact with &!: Ever in ,_.. ., . . . . which a more supportive way. It is for these reasons that this paper proposes using knowledge-based system (KBS) techniques to produce a new type of PM tool. KBS techniques are now being used successfully in many applications. A new PM tool which integrates existing PM software with a KBSt____l _I__:_:_- -__--_..L C__l _-..,A -___.: _Zua5f;u :_r____c:.._ IIILCI~CLIV~: uecwon-suppork tout cuuiu yruvtue the following benefits to an organization: e Effective PM: The expertise of good project managers could, to a degree, be replicated in the KBS, so that good management practice was applied consistently throughout an organization. The KBS could propose full evaluated alternative strategies, L---J uaseu on previousiy vaiiciated sources such as the organization’s managers. A manager could be confident of the advice that the KBS offered, because he/she could request explanations for its recommendations, and thus explore options in a constructive manner. l Enforcement of standards: The KBS could incoporate formal and informal standards, so that companywide enforcement of specific standards and methodologies could be achieved. This should happen during the three stages of project planning: initial plan creation, plan monitoring and replanning.

0263-7863/92/040213-06

@ 1992 Butterworth-Heinemann

Ltd

213

Towards an intelligent planning system a Reduced workload: A KBS could reduce the workload of project managers, freeing time for them tc concentrate on other criticai activities.

PM SUPPORT FOR CORPORATE

goals, and model the project, its environment, and user tradeoffs, such as cost, time and product, where ‘product’ encompasses project-specification issues such as quality and scope.

GOALS

Table 1 shows how a company’s corporate goals and tactical activities could be supported by a PM tool. For project-based organizations, it is the.quality of the PM tool which facilitates tactical activltles such as cost management and resource management, and these in turn determine the achievement of corporate goals. To obtain the required level of quality, the PM tool must be able to cover three knowledge-intensive stages of planning. These are as follows: Initial planning: This stage requires knowledge of factors such as nlnnninp methodolnw. marker ____‘____-_-D,, l---**‘----P sectors, resources, activities, constraints and risks. Plan monitoring: This stage involves not only merely inputting cost and progress details, but also assessing and evaluating results and interpreting symptoms. Replanning: This stage involves amending the plan efficiently without replanning from scratch. Options for fulfilling goals must be considered. The more the project manager could inform the system about project details, such as corporate goals and objectives, the better it could support him/her. The system could achieve this by assisting with initial plan creation, highlighting problems, and recommending appropriate options for consideration. Because a KBS can perform these important ‘high-value’ functions, it __.___1_ Al__ __I_ 01 _I convf2nLwIial ___-.__r:..__l _1_--:-..___.^_.. GXCG~US LII~ role ~I~~IIIIII’~ S~SLGIIIS, which provide reasonable ‘low-value’ functions such as CPA and reporting. However, conventional planning systems are mainly calculation-based, and managers must make their own interpretations. To suggest viable alternatives, a KBS can model how constraints impact

ROLE OF KBS PLANNING

Figure 1 compares the role of conventional and KBS PM tools. Most conventional planning systems permit the input of resources, tasks, durations, progress, costing data and plan amendments. The systems perform CPA, and output the results in the form of sophisticated graphs and charts. Because of the impoverished project model, the role of the system is to act as a calculation-support tool which is capable of performing limited analysis, and which provides no interactive decision support for the user. The inputs shown in Figure 1 under ‘KBS’ are qualitatively different from those listed under ‘Conventional system’. Sample plans are a case in point. These are template sections for building complete project plans. A typical sample plan includes generic action knowledge (similar to conventional ‘tasks’), which shows the actions’ minimum, maximum and average durations, the objectives they fulfil, how they are constrained, what alternative actions could be used to achieve the objectives, which resource types

Conventional

KBS

system

I I Product

goals

Meet profit targets Meet agreed quality standards Reduce costs Maximize use of resources

Tactical

activities

Cost management Resource management Quality management Production control Client management

I

knowledae

II Monitoring-‘Idetails

Table 1. Achieving corporate goals Corporate

TOOL

Specific

Facilities PM tool

required

of

Initial planning 0 scope project l allocate budget 0 set timescales l select tasks l specify standards 0 allocate resources 0 plan contingency l analyse risks

Analysis

Plan monitoring Advice

0 collect progress 0 assess impacts Replanning l investigate

alternatives l evaluate tradeoffs l modify plan efficiently

214

u

Calculation-support

Figure 1. Roles systems

Decision-

tool

of conventional

International

Journal

support

tool

and KBS planning

of Project

Management

S SMITH

have to be used to carry them out, and what the conditions are for linking them together. The benefit of this approach is that it provides a way of communicating standards and distributing acquired knowledge. A KBS-based PM tool can incorporate the requisite knowledge to act as a decision-support tool for the project manager. A KBS needs more inputs to obtain a richer model of the project and its environment. In addition to the inputs provided bv conventional svstems. can accept and*apply knowledge about the foiowingI

corporate obiectives, project objectives, resource tvnes. -
l l l

Set up initial company knowledge. Set up initial decision-support knowledge. Enter specific ‘knowledge-rich’ plans.

Each of these steps is explained in the following sections. A discussion on how to acquire and represent the different types of knowledge is included in the Approach section.

SETTING UP GENERAL KNOWLEDGE An experienced amolmt -__ _-----

The quality of the project model permits qualitatively enhanced outputs. This enables the system to store the intricate relationships between the items listed above, so that the impact of changing any one can be assessed. This also means nnnhicticatd annlvcic ran be ___--__- that ----_ __r_____ __I__ - -__-_,--performed to give warnings of potential difficulties. If a problem occurs during the plan-monitoring stage, the system can offer suggestions as to how to amend the plan, taking into account the project objectives, and tradeoffs such as time, cost and product. Thus the KBS functions not as a calculation-support tool, but as a decision-support tool. To use such a PM tool, an authorized user needs to take the following steps:

project

COMPANY

manager

of -- information ----__--- -_____ ahold ----_

knows

an enormous

the .._, his/her ____.____cnmnenv --___r-__~ ____ WRV

works. A KBS PM tool would need some of this information to enable it to provide appropriate support for a particular company. Figure 2 shows the main types of knowledge which an authorized user must set up in the system’s knowledge base. The user could be a project manager, project leader, or system administrator, as appropriate for the organization. Whatever his/her background, the user would need annrnnriate m~rr--r-----

allthnri;rntinn ---_---_I__.__.

knowledge in the system. consists of the following:

tn .-

inmlt "'r-‘,

The

am_e~d

system

0~

delete

knowledge

such as the maximization 0 Corporate goals, resource utilization. l Corporate information, which includes

of

o the resource types that the organization deals with, o the action (or pro&?&es) that it performs,

KBS PM tool Knowledge

base

I

General company knowledge

@a

Decision-support knowledge

Specific plan knowledge

aa@J=

aa

a

em

BC5%&

Figure 2. Proposed knowledge base Vol 10 No 4 November

1992

215

Towards an inteiligent planning system o the products (such as deliverables

or services) that

it provides. l

l

Sample plans can be entered to provide a library of typical plans, to ease later input, and to give&the system a basis for making suggestions. This is a means of incorporating the organization’s planning methodology. Constraints and standards can be explicitly entered so that the system can enforce company policy at all levels.

Setting up this general information &API~hlP ;n;t;o1 .TkUC‘U"I~‘,11L1(11

Pffr,d hnnn,.c2.a cIII"It) "tibc$UJcI

data, as is commonly systems; it represents environment. A KBS manager during this facilities.

will require

con-

:+ :e "r\+ ma.va1.r "r.mc.r:P II Ia ll"L Lllkls.ly I,UU,LaI1k,

found in conventional planning a model of the project manager’s PM tool would assist the project process with easy-to-use .menu

SETTING UP DECISION-SUPPORT KNOWLEDGE In addition to the factual company information, the system would also need to hold decision-support knowledge (see Figure 2). This would normally be set up by a system developer trained in using KBS techniques who had access to the source code, but amendments could be made by the end user; different users would have different authorization levels for changing the knowledge base. Three main types of decision-support knowledge are needed:

plans (see Figure 2). A special external interface would allow existing plans from a conventional planning system to be fed in, and plan data to be fed back out for display. This would mean that existing plans would not have to be reentered from scratch, and that conventional planning systems could continue to contribute in areas where they excelled, such as report production. However, as conventional plans are mainly activity networks, additional information, such as the connections to the goals, actions, resources and product specifications, would have to be set up by the user, to permit the KBS to function to the best of its ability. Alternatively, the user interface could permit the T-h.0 1lrPr f=ntpr tntn1117 brew y.wsu. nllnr UYYl tn C” V..L”, L 11u ;nn,,tc In‘pu.ir ;nc1**rln .llr.U”\, ‘“‘U”J information shown in Figure 2, and, as output, the system provides analysis to identify potential problems, as well as options for resolving these problems. Because the system can evaluate the tradeoffs between time, cost and product, it can take account of these preferences when making its recommendations. Plan details must be as full and accurate as possible, and so it is important to provide a user-friendly ;nterfare “.I”IICs”U.

A .rr&wlnnr I”,l.““II, I1

;mn LG”II,

General examples of such knowledge are facilities for checklist creation and topic prioritization. A specific ‘heuristic’ (rule of thumb) might be to produce a plan which phased 2 -1. uenverabies across the iength of a project to maximize cash flow for the organization. It is important that the initial plan be as accurate and reasonable as possible, as many projects fail because the estimates in the initial plan cannot be met. Often, detailed planning must be left until later, at which point the resourcing levels must be rechecked to ensure that the plan is still viable. Plarz-monitoring knowledge: An example is the noticing of a slippage, and the warning of the effects of this, such as the impact on deliverables, resources and future planned activities. For example, the system could flag the nonarrival of goods ordered. Plan-recovery knowledge: For example, if a resource becomes unavailable, there may be several options to repair the plan, such as that of using a substitute resource of the same type, or making the required product using an alternative process. In the case of ordered goods not arriving on time, one option would be to suggest an alternative supplier (which the system could list, using company knowledge already set up, such as preferred suppliers).

ENTERING PLANS Once the generic company and decision-making knowledge had been entered, it would be possible to enter 216

.xs4 r\r\;*ta. CallLl y”Mrw

/wrlrrn\ \??MKj

KBS ARCHITECTURE Figure 3 shows the proposed system architecture. The system incorporates

P~an-~u~~d~ngknowledge:

mPn,, lt‘vllia

interface can achieve this, together with hypertext facilities if the user needs to find out additional information about a particular item. It may also be possible to input generic plans using data from existing planning systems, but, again, these would have to be enhanced to be usable by the KBS.

three interfaces, as follows:

User interface: This can be tailored for different types of user (e.g. project manager, project leader or system administrator). It can also be flexible enough to handie from novice to experienced users in each class. Developer interface: This is for the initial system developer or for a nominated trained KBS developer within the end-user organization. External interface: This is for linking the tool with existing planning systems. The knowledge base is the store which holds items such as general company information and specific project plans. The control box specifies when and how decision-making logic is applied. Decision-support information is stored in software processes called ‘managers’; there are three of these, to handle products, actions and resources. These are analogous to a company’s production manager, project manager and line manager. An overall manager can schedule operations and handle conflicts in such areas as time, costs and product tradeoffs.

The tool would have planning components

as follows:

Knowledge builder: This would help the user set up general company knowledge and decision-support knowledge. Plan builder: This would help the project manager to set up plans. Plan monitor: This would monitor plans.

International

Journal of Project management

S SMITH

Planning

Overall

:

manager

Resource manager

manager

components

Product Developer l/F

User I/F

I

-i

/

Sample

\

/

Actual

\

I

V//A

V///l

Working memory

External

I/F

Figure 3. Proposed KBS ~rc~i~ec~~re

l

Plan recovery: This would suggest alternatives replanning.

for

These modules represent the basic building blocks of the system, and would make use of information in the knowledge base. Further components, such as a risk analyser, could be built as required, so that an incremental approach could be taken to improving PM support.

l

There may be people issues; for example, the expert may feel threatened, especially if inconsistencies or absences of knowledge are revealed.

However, if the project is scoped sufficiently well, and if appropriate techniques are used, it should be feasible to build the knowledge base. The approach which would need to be used to elicit planning knowledge includes the following activities:

As with the majority of KBS applications, a key difficulty lies in the methodology, tools and techniques for eliciting expert knowledge. Typical difficulties include the following:

Assess all relevant existing forms of knowledge about how planning is performed in an organization, for example tools, manuals, standards, project plans and material used for training. Acquire detailed knowledge from the accumulated experience of experts using techniques such as participative workshops and formal and informal interviews. Animate the knowledge acquired during the previous two steps by means of prototype systems that explore specific issues regarding the integration and use of the selected knowledge areas.

Experts may not be able to abstract their knowledge of planning. However, they tend to be good at describing examples of previous planning successes or failures. The experts can disagree.

Appropriate knowledge representations based on previous work’ are summarized in Table 2. It would not be advisable to build a system of the complexity described in this paper by only using a rule-based system (RBS). This is because an RBS lacks theexpressiveness required.

APPROACH This section considers knowledge-acquisition and -representation issues, and deals with the feasibility of the proposed system. Knowledge acquisition and repr~entation

l

l

Vol 10 No 4 November

1992

217

Towards an intelligent planning system Table 2. Knowledge representation

Knowledge

Characteristics

Knowledge type

Conceptual model

Typical

Actions Resources Products Plans Goals Tradeoffs

Can be classified hierarchically

Factual

Object orientation, which includes entities and their relationships

Object-oriented programming environment

Plan changes

Can cause impact

Factual

Triggers, which describe actions to take when an event occurs

Demons (sections of software which fire when an event occurs) and rules

Assess impact of potential change

Strategic

Simulation

Entity-relationship navigation, worlds,

Extensive knowledge of objectives, constraints, and resources

Strategic

throughout the plan

What-ifs

Plan building Plan monitoring Plan-recovery strategies

Planning knowledge, as shown in Table 2, is a combination of rich and complex factual knowledge, as well as decision-making (strategic) processes. An RBS addresses the latter, and not the former. By using a mixture of object orientation and rules, the following scenario can be handled. A plan-recovery strategy (rule) may say ‘IF a resource is unavailable, THEN suggest an available resource with the same properties as an alternative’. The system could then address the following query to the knowledge base: ‘is there a suitable alternative resource?‘.

Knowledge ‘IF


The key issues concerning a production proposed system are as follows:

of

Rules Actions can be modelled by methods (software attached to objects which can be invoked to cause an action)

A project manager would thus greatly benefit from an intelligent KBS assistant, which could ultimately help in meeting the organization’s corporate goals. REFERENCES Final Report’ Alvey PLANIT Club, UK (1988) (held on behalf of the club by SD-Scicon Ltd)

‘PLANIT

version of the

Performance

and sizing: A sizing operation would be necessary to ensure that tools were available to provide online access to the required volumes of information. Change management: Knowledge needs to be modelled, and facilities provided, to allow users to change local information as well as corporate knowledge over time.

SUMMARY A KBS PM tool such as the one proposed would have a qualitatively enhanced model of a project and its environment. This would provide considerable support to organizations in which project planning was critical to business success. The capturing of corporate information and project-planning knowledge can ensure:

218

THEN

etc.

effective PM, consistent adherence to standards and encouragement in the use of guidelines, reduced workloads for project managers, by providing interactive analysis and decision-support facilities.

1 Feasibility of proposed system

typically

implementation

Smith graduated with Sharon honours from St Andrews University, UK, in 1979. After seven years’ experience in real-time software, she completed an MSc in intelligent systems at Brunel University, UK, researching planning techniques for real-time reactive resource allocation. She has been involved with the management, specification, design, and development of AI systems for over six years. Sharon Smith was a member of the implementation team on the Alvey PLANIT project, which researched KBS techniques and developed a prototype expert system for project planning, process planning, and job-shop scheduling. She has implemented and managed KBS systems for a number of blue-chip clients, and has published work on resource allocation, planning, and design for testability, in the UK and elsewhere. She is the chairperson of the UK Association of Project Managers KBS interest group.

International

Journal of Project Management