Skip to main content

Time-Cost Tradeoff

Multi-Mode RCPSP · Crashing · NP-Hard

Construction · Schedule & Sequence · Tactical

When a project is running behind, contractors face a classic decision: which activities should be crashed — executed faster at higher cost — to meet the deadline with the minimum extra spend? Kelley (1961) formalised this as the time-cost tradeoff problem (TCTP): each activity has a piecewise-linear cost-versus-duration curve, and an LP traces the entire Pareto frontier. The discrete multi-mode RCPSP (MRCPSP) — where each activity has 2–3 named modes (normal, expedited, crashed) — is the modern discrete variant, strongly NP-hard and a first-class member of the Hartmann & Briskorn variant landscape.

Where This Decision Fits

Schedule recovery — the highlighted step is what this page optimises

Baseline RCPSPFeasible schedule
Deadline PressureTarget T < Cmax
Time-Cost TradeoffPick crash modes
Resource LevelingSmooth the peak
ExecuteMonitor & rebaseline

The Problem

Minimum-cost schedule compression under a deadline

Each activity \( i \in \{1, \ldots, n\} \) can be executed in one of \( M_i \) modes. Mode \( m \) has duration \( p_{im} \) and fixed cost \( c_{im} \); shorter-duration modes cost more (crash cost > normal cost). The planner chooses binary variables \( x_{im} \in \{0,1\} \) with \( \sum_m x_{im} = 1 \) (exactly one mode per activity) and start times \( s_i \ge 0 \). Precedence and renewable-resource constraints carry over from base RCPSP; resource demands \( r_{imk} \) now depend on mode too (a crashed activity typically uses more crews per day).

The standard objective is minimise total cost subject to a project deadline \( T \):

MRCPSP — deadline variant (cost minimisation) $$ \min \; \sum_{i=1}^{n} \sum_{m=1}^{M_i} c_{im}\, x_{im} $$ $$ \text{s.t.} \quad \sum_{m=1}^{M_i} x_{im} = 1 \qquad \forall i $$ $$ s_j \;\ge\; s_i + \sum_{m=1}^{M_i} p_{im}\, x_{im} \qquad \forall (i,j) \in \mathcal{E} $$ $$ \sum_{i\, :\, s_i \le t < s_i + p_{i\,m(i)}}\; r_{i\,m(i)\,k} \;\le\; R_k \qquad \forall k, \; \forall t $$ $$ s_{n+1} \;\le\; T, \qquad x_{im} \in \{0,1\}, \qquad s_i \ge 0 $$

Setting \( T \) to different values and re-solving traces the time-cost Pareto front: the set of non-dominated (duration, cost) pairs. Kelley (1961) showed that when cost curves are continuous and piecewise-linear convex in duration, the problem becomes a parametric LP solvable in closed form — the entire Pareto front emerges from a single sequence of pivots. The discrete MRCPSP is strongly NP-hard (Sprecher & Drexl 1998).

See the base RCPSP formulation (no crashing)

Try It Yourself

Pick a mode per activity; trace the Pareto front by varying the deadline

Time-Cost Tradeoff Solver

10 Activities · 3 Modes each
40 d

Tighter deadlines force more crash modes — watch the minimum cost rise.

Click a mode chip to pin it manually; click Solve to optimise all modes.

Each dot is the minimum cost for a given deadline. The current deadline is highlighted.

Pick a scenario, drag the deadline slider, and click Solve.

The Algorithms

LP for convex cost curves, branch-and-bound or GA for discrete multi-mode

Exact

Kelley 1961 Parametric LP

O(n m) pivots on the time-cost-compression LP

When each activity has a piecewise-linear convex cost-duration curve (a continuum of modes between normal and full crash), the entire Pareto front emerges from a single parametric LP — the critical path is identified, the cheapest crash along it is taken, the critical path is re-computed, and the process repeats until no further cost-effective compression is possible. Pedagogically elegant; the foundation every construction PM was taught as “crashing the critical path.”

Exact (discrete)

Enumeration with Priority Rules

O( product Mi · n ) — full mode-vector search

For small instances (say, up to 10 activities with 3 modes each → 59,049 mode vectors), exhaustive enumeration of mode combinations is tractable. Each mode vector is scheduled with a standard RCPSP heuristic (Serial SGS with LFT priority) to get a feasible makespan-and-cost pair. The Pareto front is extracted from the dominated set. This is what the in-page solver uses to get a provably optimal cost for each target deadline.

Metaheuristic

Multi-mode Genetic Algorithm

Iterative · population on (mode-vector, activity-list) pairs

Hartmann (2001) introduced a two-part chromosome GA for MRCPSP: one half is the activity list (priority order), the other half is the mode vector. Standard precedence-preserving crossover and mode-swap mutation. Scales to the PSPLIB MMLIB benchmark suite (up to 30 activities, 3 modes) with <1% gap to optimum on average.

Real-World Complexity

Why construction crashing goes beyond textbook TCTP

Beyond Textbook TCTP

  • Indirect costs — Overheads, site supervision, and owner penalties add a decreasing cost function in total duration that the classical TCTP objective doesn't capture directly. The combined cost curve is U-shaped.
  • Non-convex cost curves — Discrete modes with step changes (single-shift → double-shift) violate the convexity assumption that makes Kelley's LP exact. In practice, the discrete MRCPSP is the right model.
  • Stochastic durations — Crashed modes are riskier: higher variance in realised durations. Robust TCTP incorporates this.
  • Quality penalties — Some crashing reduces quality (rushed concrete cure times, compressed inspections); the Time-Cost-Quality Tradeoff (TCQT) is a three-axis extension (Babu & Suresh 1996).
  • Contractor incentives — Early-completion bonuses and late penalties per day create a staircase objective rather than a monotone one; the optimal schedule may deliberately sit at an inflection point of the bonus curve.
  • Cascading crash effects — Crashing activity A may open up a new critical path through B; the LP handles this automatically, but discrete-mode heuristics must re-evaluate after every decision.

Related RCPSP Variants

TCTP is one of >150 variants catalogued by Hartmann & Briskorn (2022)

RCPSP — base, single mode. The canonical parent problem this page extends. Fix each activity to its normal mode and deadline-cost disappears. → Open Project Scheduling
Stochastic RCPSP. Crashed durations are not just shorter, they're riskier. Stochastic variants insert time buffers rather than cost. → Open Stochastic RCPSP
Resource Leveling. Once the TCTP schedule is chosen, level the crew histogram within float. Complementary post-processing step. → Open Resource Leveling
Linear / LOB Scheduling. Highway and high-rise projects with repetitive work: the time-cost curve looks different because the bottleneck is crew continuity, not critical-path duration. → Open Linear Scheduling
Cash-Flow Optimisation. TCTP minimises undiscounted cost; cash-flow accounts for the time-value of money through interest and billing milestones. → Open Cash Flow
TCQT — Three-axis tradeoff. Babu & Suresh (1996) extend TCTP with quality as a third axis. Frontier research; not yet modelled here.

Key References

Cited above · DOIs & permanent URLs

Kelley, J. E. (1961).
“Critical-path planning and scheduling: mathematical basis.”
Operations Research, 9(3), 296–320. (Original parametric LP for continuous TCTP.) doi:10.1287/opre.9.3.296
Kelley, J. E., & Walker, M. R. (1959).
“Critical-path planning and scheduling.”
Proceedings of the Eastern Joint Computer Conference, 160–173. (CPM origin at DuPont — the underpinning of all time-cost analysis.)
De, P., Dunne, E. J., Ghosh, J. B., & Wells, C. E. (1995).
“The discrete time-cost tradeoff problem revisited.”
European Journal of Operational Research, 81(2), 225–238. (Complexity results and exact algorithms for discrete-mode TCTP.) doi:10.1016/0377-2217(93)E0334-T
Feng, C. W., Liu, L., & Burns, S. A. (1997).
“Using genetic algorithms to solve construction time-cost trade-off problems.”
Journal of Computing in Civil Engineering, 11(3), 184–189. (Early GA applied to TCTP in construction.) doi:10.1061/(ASCE)0887-3801(1997)11:3(184)
Sprecher, A., & Drexl, A. (1998).
“Multi-mode resource-constrained project scheduling by a simple, general and powerful sequencing algorithm.”
European Journal of Operational Research, 107(2), 431–450. (Establishes MRCPSP NP-hardness even for small mode counts.) doi:10.1016/S0377-2217(97)00348-2
Hartmann, S. (2001).
“Project scheduling with multiple modes: a genetic algorithm.”
Annals of Operations Research, 102(1), 111–135. (The two-part-chromosome GA referenced on this page.) doi:10.1023/A:1010902015091
Demeulemeester, E., De Reyck, B., Foubert, B., Herroelen, W., & Vanhoucke, M. (1998).
“New computational results on the discrete time-cost tradeoff problem in project networks.”
Journal of the Operational Research Society, 49(11), 1153–1163. doi:10.1057/palgrave.jors.2600634
Babu, A. J. G., & Suresh, N. (1996).
“Project management with time, cost and quality considerations.”
European Journal of Operational Research, 88(2), 320–327. (TCQT — quality as a third axis.) doi:10.1016/0377-2217(94)00202-9
Hartmann, S., & Briskorn, D. (2022).
“An updated survey of variants and extensions of the resource-constrained project scheduling problem.”
European Journal of Operational Research, 297(1), 1–14. (Most recent cataloguing of MRCPSP and TCTP variants.) doi:10.1016/j.ejor.2021.05.004

Need to optimise your construction schedules?

From time-cost tradeoff to resource-constrained scheduling and cash-flow planning, mathematical modelling can transform construction recovery decisions. Let's discuss how Operations Research can work for you.

Disclaimer
Activity names, durations, mode options, and costs shown are illustrative values chosen to demonstrate the time-cost tradeoff algorithms. Real construction projects involve hundreds of activities and far richer mode sets.
Back